Upload folder using huggingface_hub

thesis_repro/README.md

@@ -0,0 +1,28 @@
+# Thesis reproduction snapshot (Hugging Face)
+
+This folder supplements the public GitHub repository with **small, path-sensitive outputs** that are
+normally under `phylo_msa1/outputs/` (gitignored for size) and **Feng sequence benchmark**
+aggregated results (also gitignored by pattern).
+
+## Layout
+
+- `phylo_msa1_outputs/`
+  - `VEP_ALL_BASE_MODELS_STRICT_METRICS.csv` — registry-wide strict-247 VEP metrics (rebuildable from `vep_all_base_models/*.parquet`).
+  - `VEP_ALL_BASE_MODELS_STRICTV2_MATCHED_METRICS*.csv` — strict-v2 matched cohort leaderboard inputs used by `thesis_manu_v2_6/scripts/gen_strictv2_ci_artifacts.py`.
+  - `vep_all_base_models/*.parquet` — per-checkpoint score vectors for strict-247 (small).
+- `dna_foundation_benchmark_results_final/` — copy of `downstream_tasks/dna_foundation_benchmark/results_final/` used by Feng benchmark LaTeX generators.
+
+## Restore into a fresh clone
+
+From repository root after `hf download .../pevo-msa-grch38-19way`:
+
+```bash
+mkdir -p phylo_msa1/outputs
+cp -a thesis_repro/phylo_msa1_outputs/VEP_*.csv phylo_msa1/outputs/
+mkdir -p phylo_msa1/outputs/vep_all_base_models
+cp -a thesis_repro/phylo_msa1_outputs/vep_all_base_models/* phylo_msa1/outputs/vep_all_base_models/
+mkdir -p downstream_tasks/dna_foundation_benchmark
+cp -a thesis_repro/dna_foundation_benchmark_results_final downstream_tasks/dna_foundation_benchmark/results_final
+```
+
+Model checkpoints live on the **model** Hub (`jasperyeoh2/pevo-msa-mlm-19way`, `base_models/`). Training windows / full MSA corpus may require the dataset Hub snapshot + paths documented in the root `REPRODUCE.md`.

thesis_repro/dna_foundation_benchmark_results_final/baseline/enhancer_rf.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4152544212734215
+AUC,0.7708999999999999
+F1-Score,0.7125307125307125
+Accuracy,0.7075

thesis_repro/dna_foundation_benchmark_results_final/baseline/promoter_notata_251bps_rf.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.6711847630055824
+AUC,0.9142246547708224
+F1-Score,0.8474471376998453
+Accuracy,0.8362851450077485

thesis_repro/dna_foundation_benchmark_results_final/dnabert2_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4605760803760181
+AUC,0.797125
+F1-Score,0.7365853658536585
+Accuracy,0.73

thesis_repro/dna_foundation_benchmark_results_final/dnabert2_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.7055929314759852
+AUC,0.9187466364960829
+F1-Score,0.8509201785908744
+Accuracy,0.8484613681647111

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w1024_ft128_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4700940282094033
+AUC,0.8045250000000002
+F1-Score,0.7376237623762376
+Accuracy,0.735

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w1024_ft128_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.5026643677346545
+AUC,0.8273264086212394
+F1-Score,0.765142797363741
+Accuracy,0.7514943546601727

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w256_ft126_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.48410814469322405
+AUC,0.81395
+F1-Score,0.755868544600939
+Accuracy,0.74

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w256_ft126_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4733248853206569
+AUC,0.8076991793235673
+F1-Score,0.7462020609794965
+Accuracy,0.7355545716183307

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w512_ft123_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.5055120903686235
+AUC,0.812075
+F1-Score,0.7579462102689487
+Accuracy,0.7525

thesis_repro/dna_foundation_benchmark_results_final/phylo_l7_w512_ft123_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4612172920443689
+AUC,0.7994494905750267
+F1-Score,0.749715938436112
+Accuracy,0.7317910117334514

thesis_repro/dna_foundation_benchmark_results_final/phylo_msa_exp22_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4422166387140533
+AUC,0.7837500000000001
+F1-Score,0.7052631578947368
+Accuracy,0.72

thesis_repro/dna_foundation_benchmark_results_final/phylo_msa_exp22_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.3957373414715589
+AUC,0.7588808234771399
+F1-Score,0.7039190353143842
+Accuracy,0.6955944210759354

thesis_repro/dna_foundation_benchmark_results_final/phylo_n10_exp206_meanpool/enhancer_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4705886038624533
+AUC,0.8046
+F1-Score,0.7282051282051282
+Accuracy,0.735

thesis_repro/dna_foundation_benchmark_results_final/phylo_n10_exp206_meanpool/promoter_notata_251bps_elastic.csv

@@ -0,0 +1,5 @@
+Metric,Value
+MCC,0.4736215562521326
+AUC,0.8042027405934882
+F1-Score,0.7513083525225036
+Accuracy,0.7369935798096081

thesis_repro/phylo_msa1_outputs/VEP_ALL_BASE_MODELS_STRICTV2_MATCHED_METRICS.csv

@@ -0,0 +1,176 @@
+ablation_id,in_ablation_registry,class,configuration,checkpoint_stem,n,pos,neg,auroc,auprc,auroc_flipped,auprc_flipped,auroc_calibrated,auprc_calibrated,polarity,score_mean,score_std,branch_tag
+1,True,Baseline,"Baseline (Full Configuration: MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_01,2000,1000,1000,0.44854049999999995,0.4559150461813925,0.55145949999999999,0.52914947309059168,0.55145949999999999,0.52914947309059168,flipped,-1.715979248046875,0.58755854204363589,legacy 7-way line
+2,True,Component Removal,"Component Removal: w/o Flip Nonconserved (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_02,2000,1000,1000,0.43929250000000003,0.44985023101858201,0.56070750000000003,0.53171186696418427,0.56070750000000003,0.53171186696418427,flipped,-1.67484375,0.563917349867052,legacy 7-way line
+3,True,Component Removal,"Component Removal: w/o Auxiliary Features (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_03,2000,1000,1000,0.51639800000000002,0.49877893367748338,0.48360199999999998,0.47262231395750609,0.51639800000000002,0.49877893367748338,raw,-1.69120703125,0.46584178244219454,legacy 7-way line
+4,True,Component Removal,"Component Removal: w/o Soft Masked Loss (MLM 0.15, LR 0.0002, 512/8/6)",full_training_04,2000,1000,1000,0.43448999999999993,0.44241468941242795,0.56550999999999996,0.52582287366250302,0.56550999999999996,0.52582287366250302,flipped,-1.8489296875000001,0.60380996910475337,legacy 7-way line
+5,True,Component Removal,"Component Removal: w/o Conservation Weight (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_05,2000,1000,1000,0.42124800000000007,0.43484812892155073,0.57875199999999993,0.54366184988685251,0.57875199999999993,0.54366184988685251,flipped,-1.8089472656250001,0.55679599407539027,legacy 7-way line
+6,True,Hyperparameter: MLM,"Hyperparameter: MLM 0.1 (Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_06,2000,1000,1000,0.43224400000000007,0.44506912024168488,0.56775599999999993,0.54855174385158856,0.56775599999999993,0.54855174385158856,flipped,-1.7975959472656251,0.54447818633236922,legacy 7-way line
+7,True,Hyperparameter: MLM,"Hyperparameter: MLM 0.2 (Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_07,2000,1000,1000,0.43174600000000002,0.43935256547718865,0.56825400000000004,0.5296417918916938,0.56825400000000004,0.5296417918916938,flipped,-1.7247529296874999,0.54390337338934802,legacy 7-way line
+8,True,Hyperparameter: Soft Mask,"Hyperparameter: Soft Mask 0.05 (MLM 0.15, LR 0.0002, 512/8/6)",full_training_08,2000,1000,1000,0.42351749999999994,0.43961361981471497,0.57648250000000001,0.53991048322842516,0.57648250000000001,0.53991048322842516,flipped,-1.7354706420898438,0.59709798024255412,legacy 7-way line
+9,True,Hyperparameter: Soft Mask,"Hyperparameter: Soft Mask 0.2 (MLM 0.15, LR 0.0002, 512/8/6)",full_training_09,2000,1000,1000,0.43735150000000006,0.4465957355517628,0.56264850000000011,0.53014818656767926,0.56264850000000011,0.53014818656767926,flipped,-1.6657431640625,0.61752790824974713,legacy 7-way line
+10,True,Hyperparameter: Learning Rate,"Hyperparameter: Learning Rate 0.0001 (MLM 0.15, Soft Mask 0.1, 512/8/6)",full_training_10,2000,1000,1000,0.42732199999999998,0.44089130330192672,0.57267800000000002,0.53080579335174938,0.57267800000000002,0.53080579335174938,flipped,-1.7955468750000001,0.53547679096971235,legacy 7-way line
+100,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.085: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_100_64bp_MLM_0.085,2000,1000,1000,0.43207500000000004,0.44418716094542743,0.56792500000000001,0.54069185802026309,0.56792500000000001,0.54069185802026309,flipped,-1.9959218750000001,1.1251560836276957,legacy 7-way line
+101,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + LR=0.00018: MLM 0.08, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_101_64bp_LR_0.00018,2000,1000,1000,0.38069700000000006,0.40851423456033381,0.61930300000000016,0.56795890998822296,0.61930300000000016,0.56795890998822296,flipped,-1.9313515625,0.73528603028051853,legacy 7-way line
+102,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + LR=0.00022: MLM 0.08, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_102_64bp_LR_0.00022,2000,1000,1000,0.38706049999999997,0.41140833795550186,0.61293949999999997,0.56150538949427053,0.61293949999999997,0.56150538949427053,flipped,-2.0171406250000001,0.73201684067328454,legacy 7-way line
+103,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=750: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_103_64bp_warmup_750,2000,1000,1000,0.38235750000000007,0.40681549522430693,0.61764249999999998,0.55484458737671583,0.61764249999999998,0.55484458737671583,flipped,-1.9926523437500001,0.67194240031434982,legacy 7-way line
+104,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=1250: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_104_64bp_warmup_1250,2000,1000,1000,0.39745849999999999,0.41815993640522775,0.60254149999999995,0.55911235576264673,0.60254149999999995,0.55911235576264673,flipped,-2.190541015625,1.0692364318554053,legacy 7-way line
+105,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=1500: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_105_64bp_warmup_1500,2000,1000,1000,0.38422000000000006,0.40749726224242871,0.61577999999999999,0.55190407522337703,0.61577999999999999,0.55190407522337703,flipped,-2.0982617187499999,0.75091561922539296,legacy 7-way line
+106,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + soft_masked_loss_weight=0.15: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_106_64bp_soft_mask_0.15,2000,1000,1000,0.39477600000000002,0.41397721965106826,0.60522399999999998,0.55735278956134438,0.60522399999999998,0.55735278956134438,flipped,-1.95669140625,0.69090450578756091,legacy 7-way line
+107,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 类别权重Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_107_64bp_class_weights,2000,1000,1000,0.38668150000000007,0.41468975447592321,0.61331849999999999,0.57604430506783455,0.61331849999999999,0.57604430506783455,flipped,-2.0939804687499999,0.94449902465673019,legacy 7-way line
+108,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True (alpha=0.25, gamma=2.0), eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_108_64bp_focal_loss,2000,1000,1000,0.37609700000000001,0.40808469571825745,0.62390299999999999,0.57352602451566703,0.62390299999999999,0.57352602451566703,flipped,-1.366015625,0.56282192936125175,legacy 7-way line
+109,True,Model Capacity: 64bp,"Exp 80优化 + 64bp窗口 + 768/12/8架构: MLM 0.08, LR 0.0002, cosine, warmup 1000, 768/12/8, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_109_64bp_768_12_8,2000,1000,1000,0.42178650000000001,0.42662862809384211,0.57821349999999994,0.53964387562743876,0.57821349999999994,0.53964387562743876,flipped,-1.7554882812499999,0.59820996580861896,legacy 7-way line
+11,True,Hyperparameter: Learning Rate,"Hyperparameter: Learning Rate 0.0003 (MLM 0.15, Soft Mask 0.1, 512/8/6)",full_training_11,2000,1000,1000,0.44316449999999996,0.45218345371864471,0.55683550000000004,0.54368933927974006,0.55683550000000004,0.54368933927974006,flipped,-1.6347631835937499,0.66517304491790274,legacy 7-way line
+110,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 更激进类别权重Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True (aggressive), eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_110_64bp_class_weights_aggressive,2000,1000,1000,0.38901349999999996,0.41293444989017591,0.61098649999999988,0.56783444460906574,0.61098649999999988,0.56783444460906574,flipped,-2.1781669921875002,1.0146229251893464,legacy 7-way line
+111,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss (alpha=0.5, gamma=1.5): MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_111_64bp_focal_loss_tuned,2000,1000,1000,0.37646050000000003,0.40578971953390863,0.62353950000000002,0.563950491321207,0.62353950000000002,0.563950491321207,flipped,-1.4062382812500001,0.56627889181033098,legacy 7-way line
+112,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 组合优化（类别权重 + Focal Loss）: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_112_64bp_combined_loss,2000,1000,1000,0.38206699999999993,0.40699069062815485,0.61793300000000007,0.55672269971587729,0.61793300000000007,0.55672269971587729,flipped,-1.4761367187500001,0.70471303003183883,legacy 7-way line
+113,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss (alpha=0.25, gamma=2.5): MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_113_64bp_focal_loss_gamma25,2000,1000,1000,0.38611449999999997,0.41694465774221889,0.61388549999999997,0.57650839169819479,0.61388549999999997,0.57650839169819479,flipped,-1.1071015625,0.46286575744120573,legacy 7-way line
+114,True,GPN-MSA: Original Implementation,"GPN-MSA (Original Implementation): MLM 0.15, LR 0.0001, cosine scheduler, warmup 1000, 512/8/6, n_aux=30, batch_size=512×1",full_training_114_64bp_MLM_0.08_retrain,2000,1000,1000,0.43051699999999998,0.44150531436118678,0.56948300000000018,0.54511997732470729,0.56948300000000018,0.54511997732470729,flipped,-1.8644238281250001,1.0153669890402726,legacy 7-way line
+115,True,Masking Ratio,"Masking Ratio: MLM 0.30 (基于Exp 10: 128bp + aux_features, LR 0.0001, batch=2048)",full_training_115,2000,1000,1000,0.39720649999999996,0.41345424264420022,0.60279349999999998,0.54728606433283145,0.60279349999999998,0.54728606433283145,flipped,-2.0321738281249999,0.67715959244748669,legacy 7-way line
+116,True,Masking Ratio,"Masking Ratio: MLM 0.50 (基于Exp 10: 128bp + aux_features, LR 0.0001, batch=2048)",full_training_116,2000,1000,1000,0.36383100000000007,0.40025089290597504,0.63616899999999998,0.59077846034340831,0.63616899999999998,0.59077846034340831,flipped,-2.1205742187499999,0.70737998899974563,legacy 7-way line
+117,True,Block/Span Masking,"Span Masking: mean_span_length=3.0 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_117,2000,1000,1000,0.39881100000000014,0.41431288297143576,0.60118899999999997,0.54438905963251349,0.60118899999999997,0.54438905963251349,flipped,-2.0158593749999998,0.6719536586404985,legacy 7-way line
+118,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_118,2000,1000,1000,0.38792349999999998,0.40895363681082253,0.61207650000000002,0.55333212318166147,0.61207650000000002,0.55333212318166147,flipped,-2.0445751953125,0.68306214260650788,legacy 7-way line
+119,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+A1 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_119,2000,1000,1000,0.38167150000000005,0.40817363090207948,0.61832849999999995,0.57021322071510838,0.61832849999999995,0.57021322071510838,flipped,-2.21384765625,0.72870260853379687,legacy 7-way line
+12,True,Architecture,"Architecture: 384/8/6 (Small Model, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_12,2000,1000,1000,0.45593600000000001,0.46693552745335271,0.54406399999999999,0.51848993043011404,0.54406399999999999,0.51848993043011404,flipped,-1.61380224609375,0.62625447214760721,legacy 7-way line
+120,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+A1+A2 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_120,2000,1000,1000,0.39058899999999996,0.41075117921262028,0.60941099999999992,0.55564537698531058,0.60941099999999992,0.55564537698531058,flipped,-2.2291210937499999,0.73219015058297376,legacy 7-way line
+121,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 64bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_121,2000,1000,1000,0.40407799999999999,0.42174784054995368,0.59592200000000006,0.57347224608702752,0.59592200000000006,0.57347224608702752,flipped,-2.0025864257812498,1.0586120423475365,legacy 7-way line
+122,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 256bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_122,2000,1000,1000,0.37030650000000004,0.40367575394663779,0.62969350000000002,0.5978725980041113,0.62969350000000002,0.5978725980041113,flipped,-2.020904296875,0.87139939939457633,legacy 7-way line
+123,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_123,2000,1000,1000,0.39001300000000005,0.41030176306269417,0.60998700000000006,0.55503819646692309,0.60998700000000006,0.55503819646692309,flipped,-2.0376074218750002,0.72311432635211803,legacy 7-way line
+124,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_124,2000,1000,1000,0.38484800000000008,0.40901517598011028,0.61515199999999992,0.56156960260270194,0.61515199999999992,0.56156960260270194,flipped,-2.04276171875,0.7466503996655186,legacy 7-way line
+125,True,Block/Span Masking,"Span Masking: mean_span_length=12.0 (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_125,2000,1000,1000,0.38008649999999994,0.40806736594686072,0.61991350000000001,0.56475775536125883,0.61991350000000001,0.56475775536125883,flipped,-1.9673828124999999,0.7414007085788038,legacy 7-way line
+126,True,Window Size: 256bp,"Window Size: 256bp (MLM 0.15, LR 0.0001, batch=2048, 512/8/12)",full_training_126,2000,1000,1000,0.38134050000000003,0.40912272622189094,0.61865949999999992,0.56651834017584857,0.61865949999999992,0.56651834017584857,flipped,-2.160326171875,0.7562099438133314,legacy window sweep
+127,True,Window Size: 512bp,"Window Size: 512bp (MLM 0.15, LR 0.0001, batch=2048, 768/12/6)",full_training_127,2000,1000,1000,0.382739,0.40896980193516236,0.61726099999999995,0.56440352029410046,0.61726099999999995,0.56440352029410046,flipped,-2.07784375,0.75054222155903616,legacy window sweep
+128,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0001, batch=128, 768/12/12)",full_training_128,2000,1000,1000,0.36644950000000004,0.40329493857628823,0.63355049999999991,0.59305408003855309,0.63355049999999991,0.59305408003855309,flipped,-1.9325117187500001,0.53567497771263928,legacy window sweep
+129,True,Window Size: 64bp,"Window Size: 64bp (MLM 0.15, LR 0.0001, batch=2048, 768/12/12)",full_training_129,2000,1000,1000,0.38053199999999998,0.40937177723800322,0.61946800000000002,0.58074278714406735,0.61946800000000002,0.58074278714406735,flipped,-2.0963906250000002,0.80015889520042172,legacy window sweep
+13,True,Architecture,"Architecture: 768/12/8 (Large Model, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_13,2000,1000,1000,0.43187050000000005,0.44728440985345996,0.56812950000000007,0.5349908635659566,0.56812950000000007,0.5349908635659566,flipped,-1.783711181640625,0.68430515191705599,legacy 7-way line
+130,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 256bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_130,2000,1000,1000,0.39439850000000004,0.41310545412983529,0.60560150000000001,0.55389716964113833,0.60560150000000001,0.55389716964113833,flipped,-2.3113125000000001,0.81642564630069048,legacy 7-way line
+131,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_131,2000,1000,1000,0.38421500000000003,0.40945656372596739,0.61578500000000003,0.56163513985583779,0.61578500000000003,0.56163513985583779,flipped,-2.29276171875,0.83989906423039473,legacy 7-way line
+132,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_132,2000,1000,1000,0.37239800000000001,0.40215162056011938,0.62760199999999999,0.56297902566827385,0.62760199999999999,0.56297902566827385,flipped,-2.31621875,0.6973991544936633,legacy 7-way line
+133,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0003, batch=2048, 512/8/6)",full_training_133,2000,1000,1000,0.39210250000000002,0.41197165300998362,0.60789749999999998,0.55399781181920948,0.60789749999999998,0.55399781181920948,flipped,-2.0212021484374998,0.73142879711555031,legacy window sweep
+134,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0001, batch=2048, 512/8/6, warmup=2000)",full_training_134,2000,1000,1000,0.388268,0.41110329906456289,0.61173199999999994,0.55949735341081597,0.61173199999999994,0.55949735341081597,flipped,-2.012318359375,0.74535795829159623,legacy window sweep
+135,True,Loss Optimization,"Focal Loss: gamma=1.0 (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_135,2000,1000,1000,0.39332250000000007,0.41419424862693088,0.60667749999999998,0.55702693473438625,0.60667749999999998,0.55702693473438625,flipped,-1.6000478515625001,0.62953640880464024,legacy 7-way line
+136,True,特定序列Masking,"Exp 136 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_136,2000,1000,1000,0.36425150000000001,0.40001943845911719,0.63574850000000005,0.58261079032121732,0.63574850000000005,0.58261079032121732,flipped,-2.1083867187499998,0.69881113720225241,legacy 7-way line
+137,True,特定序列Masking,"Exp 137 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_137,2000,1000,1000,0.38462600000000002,0.41105759916332485,0.61537399999999998,0.57219536450062913,0.61537399999999998,0.57219536450062913,flipped,-2.0949941406249999,1.0020197046116524,legacy 7-way line
+138,True,特定序列Masking,"Exp 138 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_138,2000,1000,1000,0.39088200000000001,0.41729036830950228,0.60911799999999994,0.57864322281260649,0.60911799999999994,0.57864322281260649,flipped,-1.9655859375,0.8876872584549369,legacy 7-way line
+139,True,特定序列Masking,"Exp 139 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_139,2000,1000,1000,0.38131150000000003,0.40988098859687572,0.61868849999999997,0.5787542315944223,0.61868849999999997,0.5787542315944223,flipped,-1.9416015625,0.67306319240775681,legacy 7-way line
+14,True,Optimized Combination,Optimized Combination: MLM 0.2 + Soft Mask 0.2 + LR 0.0001 (512/8/6),full_training_14,2000,1000,1000,0.43119150000000001,0.44070240577267938,0.56880850000000005,0.53129598442391468,0.56880850000000005,0.53129598442391468,flipped,-1.7762890625000001,0.60013124111420768,legacy 7-way line
+140,True,特定序列Masking,"Exp 140 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_140,2000,1000,1000,0.35899400000000004,0.3975845928991113,0.64100599999999996,0.58515561566597118,0.64100599999999996,0.58515561566597118,flipped,-2.1130859375000002,0.70410185506235257,legacy 7-way line
+141,True,特定序列Masking,"Exp 141 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_141,2000,1000,1000,0.37847399999999998,0.40904410353908782,0.62152600000000002,0.58747314092993275,0.62152600000000002,0.58747314092993275,flipped,-2.0449882812500002,0.75152023569401827,legacy 7-way line
+142,True,特定序列Masking,"Exp 142 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_142,2000,1000,1000,0.39312750000000002,0.41272594917013794,0.60687249999999993,0.55194084661389509,0.60687249999999993,0.55194084661389509,flipped,-2.1193886718749999,0.81387484123686604,legacy 7-way line
+143,True,特定序列Masking,"Exp 143 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_143,2000,1000,1000,0.42872250000000001,0.43490644159044267,0.57127749999999999,0.54101409825544644,0.57127749999999999,0.54101409825544644,flipped,-2.0679199218750002,1.0346879198816037,legacy 7-way line
+144,True,特定序列Masking,"Exp 144 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_144,2000,1000,1000,0.39303050000000006,0.41371847211520907,0.60696949999999994,0.55669958097884675,0.60696949999999994,0.55669958097884675,flipped,-1.5930400390624999,0.63124741466559331,legacy 7-way line
+145,True,Auxiliary Features数量探索,"Exp 145 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=3, 128bp",full_training_145,2000,1000,1000,0.49739699999999998,0.50977066600956833,0.50260300000000002,0.49893447932570811,0.50260300000000002,0.50977066600956833,flipped,-2.5244453125000001,1.2239317767259701,legacy 7-way line
+146,True,Auxiliary Features数量探索,"Exp 146 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=6, 128bp",full_training_146,2000,1000,1000,0.4566925,0.46172725096235334,0.54330749999999994,0.53580451016919861,0.54330749999999994,0.53580451016919861,flipped,-1.8562197265625,1.2694814122725917,legacy 7-way line
+147,True,Auxiliary Features数量探索,"Exp 147 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=9, 128bp",full_training_147,2000,1000,1000,0.4468705,0.45780423335076154,0.55312950000000005,0.55008070702755374,0.55312950000000005,0.55008070702755374,flipped,-1.785634765625,1.054050230604568,legacy 7-way line
+148,True,Auxiliary Features数量探索,"Exp 148 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=12, 128bp",full_training_148,2000,1000,1000,0.4105355,0.42844867800676556,0.58946450000000006,0.56474046204882167,0.58946450000000006,0.56474046204882167,flipped,-1.8348164062500001,0.87742631157021822,legacy 7-way line
+149,True,Auxiliary Features数量探索,"Exp 149 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_149,2000,1000,1000,0.49739699999999998,0.50977066600956833,0.50260300000000002,0.49893447932570811,0.50260300000000002,0.50977066600956833,flipped,-2.5244453125000001,1.2239317767259701,legacy 7-way line
+15,True,Optimized Combination,Optimized Combination: MLM 0.2 + Soft Mask 0.2 + LR 0.0001 (768/12/8),full_training_15,2000,1000,1000,0.41896800000000001,0.43089283290864755,0.58103199999999999,0.53701723174475025,0.58103199999999999,0.53701723174475025,flipped,-1.8625002441406251,0.83975419019937403,legacy 7-way line
+150,True,Auxiliary Features数量探索,"Exp 150 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=18, 128bp",full_training_150,2000,1000,1000,0.44553999999999999,0.44772762662543364,0.55446000000000006,0.53506909240976896,0.55446000000000006,0.53506909240976896,flipped,-1.5358415527343749,0.68356326883210372,legacy 7-way line
+151,True,Auxiliary Features数量探索,"Exp 151 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=21, 128bp",full_training_151,2000,1000,1000,0.49301099999999998,0.4862027867637102,0.50698899999999991,0.49331238685059509,0.50698899999999991,0.49331238685059509,flipped,-0.49822396999597551,0.86986948320558333,legacy 7-way line
+152,True,Auxiliary Features数量探索,"Exp 152 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=24, 128bp",full_training_152,2000,1000,1000,0.49301099999999998,0.4862027867637102,0.50698899999999991,0.49331238685059509,0.50698899999999991,0.49331238685059509,flipped,-0.49822396999597551,0.86986948320558333,legacy 7-way line
+153,True,Auxiliary Features数量探索,"Exp 153 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=27, 128bp",full_training_153,2000,1000,1000,0.49301099999999998,0.4862027867637102,0.50698899999999991,0.49331238685059509,0.50698899999999991,0.49331238685059509,flipped,-0.49822396999597551,0.86986948320558333,legacy 7-way line
+154,True,Auxiliary Features数量探索,"Exp 154 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_154,2000,1000,1000,0.4566925,0.46172725096235334,0.54330749999999994,0.53580451016919861,0.54330749999999994,0.53580451016919861,flipped,-1.8562197265625,1.2694814122725917,legacy 7-way line
+155,True,组合策略,"Exp 155 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_155,2000,1000,1000,0.42405299999999996,0.44274736700966855,0.57594699999999999,0.54360112272026295,0.57594699999999999,0.54360112272026295,flipped,-2.4811640625,0.64401210527119745,legacy 7-way line
+156,True,组合策略,"Exp 156 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_156,2000,1000,1000,0.47873450000000006,0.49637267175242444,0.52126550000000005,0.51462381728904305,0.52126550000000005,0.51462381728904305,flipped,-2.501578125,0.84904714500630607,legacy 7-way line
+157,True,组合策略,"Exp 157 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=9, 128bp",full_training_157,2000,1000,1000,0.379832,0.40712028352098689,0.62016800000000005,0.57415159579342534,0.62016800000000005,0.57415159579342534,flipped,-1.9048437499999999,0.63913362990457634,legacy 7-way line
+158,True,组合策略,"Exp 158 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=21, 128bp",full_training_158,2000,1000,1000,0.47719099999999998,0.48452081893462751,0.52280899999999997,0.50595573462371402,0.52280899999999997,0.50595573462371402,flipped,-0.47498019820451737,0.91812012901150652,legacy 7-way line
+159,True,组合策略,"Exp 159 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_159,2000,1000,1000,0.46058750000000004,0.47881680836583179,0.53941250000000007,0.52798835143279454,0.53941250000000007,0.52798835143279454,flipped,-2.5245390625000002,0.72934401708073593,legacy 7-way line
+16,True,Optimized Combination,Optimized Combination: MLM 0.1 + Soft Mask 0.2 + LR 0.0001 (512/8/6),full_training_16,2000,1000,1000,0.44283950000000005,0.45251249678540872,0.55716050000000006,0.52061004065000904,0.55716050000000006,0.52061004065000904,flipped,-1.7212302246093749,0.57157316970246619,legacy 7-way line
+160,True,组合策略,"Exp 160 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_160,2000,1000,1000,0.48100399999999999,0.48004304948454601,0.51899600000000001,0.50588154765575399,0.51899600000000001,0.50588154765575399,flipped,-2.553484375,0.56262435062136207,legacy 7-way line
+161,True,组合策略,"Exp 161 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_161,2000,1000,1000,0.44011349999999994,0.4559445898116864,0.55988650000000006,0.53058500192485125,0.55988650000000006,0.53058500192485125,flipped,-1.9087519531249999,0.62891714688431821,legacy 7-way line
+162,True,基于最佳配置的优化,"Exp 162 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_162,2000,1000,1000,0.37784000000000001,0.40494803436171922,0.62216000000000005,0.56894560135716032,0.62216000000000005,0.56894560135716032,flipped,-2.1712773437499999,0.96380660975893628,legacy 7-way line
+163,True,基于最佳配置的优化,"Exp 163 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_163,2000,1000,1000,0.469194,0.48234986157471227,0.530806,0.52861591018739218,0.530806,0.52861591018739218,flipped,-2.4595156249999999,0.66149636888264318,legacy 7-way line
+164,True,基于最佳配置的优化,"Exp 164 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_164,2000,1000,1000,0.46332400000000001,0.47197461019847986,0.53667599999999993,0.51614468415683434,0.53667599999999993,0.51614468415683434,flipped,-2.4830156250000002,0.63579321747000361,legacy 7-way line
+165,True,基于最佳配置的优化,"Exp 165 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_165,2000,1000,1000,0.39303050000000006,0.41371847211520907,0.60696949999999994,0.55669958097884675,0.60696949999999994,0.55669958097884675,flipped,-1.5930400390624999,0.63124741466559331,legacy 7-way line
+166,True,基于最佳配置的优化,"Exp 166 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_166,2000,1000,1000,0.44578600000000002,0.45905915711811685,0.5542140000000001,0.52778804808846558,0.5542140000000001,0.52778804808846558,flipped,-1.8930039062499999,0.64112045102616955,legacy 7-way line
+167,True,基于最佳配置的优化,"Exp 167 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_167,2000,1000,1000,0.42872250000000001,0.43490644159044267,0.57127749999999999,0.54101409825544644,0.57127749999999999,0.54101409825544644,flipped,-2.0679199218750002,1.0346879198816037,legacy 7-way line
+168,True,基于最佳配置的优化,"Exp 168 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_168,2000,1000,1000,0.46307299999999996,0.48160418928013426,0.53692700000000004,0.52138438308292301,0.53692700000000004,0.52138438308292301,flipped,-2.43975390625,0.86481747037442491,legacy 7-way line
+169,True,10-primates数据集,Exp 169 (see config full_training_169.json),full_training_169,2000,1000,1000,0.27601049999999994,0.36716545575277176,0.72398950000000006,0.67992830510271429,0.72398950000000006,0.67992830510271429,flipped,-6.8474414062499998,4.5008273520827506,10-primate (early)
+17,True,Architecture,"Architecture: 768/8/6 (Hidden Size Only, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_17,2000,1000,1000,0.41743800000000003,0.44230956721177339,0.58256200000000002,0.55938414273276715,0.58256200000000002,0.55938414273276715,flipped,-1.8028193359375,0.62197106730687757,legacy 7-way line
+18,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_18,2000,1000,1000,0.44482850000000002,0.46275103852998944,0.55517150000000004,0.53208732951814586,0.55517150000000004,0.53208732951814586,flipped,-1.7147823486328124,0.6787976670979734,legacy 7-way line
+19,True,Optimal Combination,Optimal Combination: MLM 0.2 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_19,2000,1000,1000,0.434137,0.44858282244666409,0.565863,0.54230586194591468,0.565863,0.54230586194591468,flipped,-1.7031787109375001,0.69378859058649944,legacy 7-way line
+20,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.1 + LR 0.0002,full_training_20,2000,1000,1000,0.40937199999999996,0.43285091140332543,0.59062799999999993,0.54893551529566365,0.59062799999999993,0.54893551529566365,flipped,-1.80390380859375,0.61458827099244429,legacy 7-way line
+201,True,10-primates新数据集: 1024bp基准,"Exp 201 (新10-primates数据集): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_201,2000,1000,1000,0.32619149999999997,0.3853566574861142,0.67380850000000003,0.62935740333557766,0.67380850000000003,0.62935740333557766,flipped,-3.6400839843749999,2.9200669632843259,new 10-primate
+202,True,10-primates新数据集: MLM 0.1,"Exp 202 (新10-primates数据集): MLM 0.1, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_202,2000,1000,1000,0.33451550000000002,0.38810869811483595,0.66548450000000003,0.62458159980927097,0.66548450000000003,0.62458159980927097,flipped,-4.0876206054687501,3.0601385517866251,new 10-primate
+203,True,10-primates新数据集: Soft Mask 0.2,"Exp 203 (新10-primates数据集): MLM 0.15, Soft Mask 0.2, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_203,2000,1000,1000,0.34327199999999997,0.39140974400764528,0.65672800000000009,0.61739865771452784,0.65672800000000009,0.61739865771452784,flipped,-3.3632358398437501,2.5507227420073444,new 10-primate
+204,True,10-primates新数据集: LR 0.0002,"Exp 204 (新10-primates数据集): MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6, n_aux=90, 1024bp",full_training_204,2000,1000,1000,0.33743300000000004,0.38939372847859599,0.66256699999999991,0.6216540233650435,0.66256699999999991,0.6216540233650435,flipped,-3.9881684570312501,2.941157637735079,new 10-primate
+206,True,10-primates新数据集,Exp 206 (see config full_training_206.json),full_training_206,2000,1000,1000,0.27480500000000002,0.36761209205410844,0.72519500000000003,0.66918093986927651,0.72519500000000003,0.66918093986927651,flipped,-6.4859648437499997,4.4431686743950394,new 10-primate
+207,True,10-primates新数据集,Exp 207 (see config full_training_207.json),full_training_207,2000,1000,1000,0.26605699999999999,0.36457961207447898,0.73394300000000001,0.6847323524773955,0.73394300000000001,0.6847323524773955,flipped,-6.8512402343750001,4.6217662821809089,new 10-primate
+209,False,Extended run,Experiment 209: checkpoint in outputs/full_training_209; not yet in ABLATION_SUMMARY.csv,full_training_209,2000,1000,1000,0.28177600000000003,0.3684777964541287,0.71822399999999997,0.66629966708085986,0.71822399999999997,0.66629966708085986,flipped,-6.76126953125,4.6764834391662573,new 10-primate
+21,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.2 + LR 0.0002,full_training_21,2000,1000,1000,0.41145800000000005,0.43533667860866126,0.5885419999999999,0.5653760134620599,0.5885419999999999,0.5653760134620599,flipped,-1.779107177734375,0.62081505460755038,legacy 7-way line
+210,False,Extended run,Experiment 210: checkpoint in outputs/full_training_210; not yet in ABLATION_SUMMARY.csv,full_training_210,2000,1000,1000,0.27377500000000005,0.36752514979823658,0.72622500000000012,0.6740753518571917,0.72622500000000012,0.6740753518571917,flipped,-5.4673437500000004,3.673101630637146,new 10-primate
+22,True,Optimal Combination,Optimal Combination: MLM 0.15 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_22,2000,1000,1000,0.42736550000000001,0.44964728239018342,0.57263449999999994,0.54788649564126046,0.57263449999999994,0.54788649564126046,flipped,-1.71562744140625,0.61624075017471358,legacy 7-way line
+23,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 (768/8/6),full_training_23,2000,1000,1000,0.42006500000000008,0.44465576300092369,0.57993500000000009,0.55970312163659164,0.57993500000000009,0.55970312163659164,flipped,-1.8161669921875001,0.66322390608820447,legacy 7-way line
+24,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.2 + LR 0.0002 (512/8/6),full_training_24,2000,1000,1000,0.42447849999999998,0.43845433408229645,0.57552150000000002,0.55022618018315472,0.57552150000000002,0.55022618018315472,flipped,-1.7752514648437501,0.59911695683052912,legacy 7-way line
+25,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.1 + LR 0.00015 (512/8/6),full_training_25,2000,1000,1000,0.42972900000000003,0.44910012607300498,0.57027099999999997,0.54232413044707684,0.57027099999999997,0.54232413044707684,flipped,-1.6911528320312501,0.65354459592586134,legacy 7-way line
+26,True,Improvement,Improvement: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 (768/8/6),full_training_26,2000,1000,1000,0.4237765,0.43940221510854549,0.5762235,0.55510728999017089,0.5762235,0.55510728999017089,flipped,-1.8021002197265625,0.60042463453206785,legacy 7-way line
+27,True,Improvement,Improvement: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 + Cosine LR (512/8/6),full_training_27,2000,1000,1000,0.44232600000000005,0.44779431609496545,0.557674,0.528488016864884,0.557674,0.528488016864884,flipped,-1.78655908203125,0.55887245745177572,legacy 7-way line
+28,True,Architecture: Heads,"Architecture: 480/12/6 (Heads Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_28,2000,1000,1000,0.43253400000000003,0.44021880676986569,0.56746600000000003,0.52838921091938373,0.56746600000000003,0.52838921091938373,flipped,-1.7774091796875,0.54617173453530954,legacy 7-way line
+29,True,Architecture: Heads,"Architecture: 512/16/6 (Heads Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_29,2000,1000,1000,0.41986750000000006,0.43363167594786134,0.58013249999999994,0.55037456880857871,0.58013249999999994,0.55037456880857871,flipped,-1.7824951171875001,0.49910787566014458,legacy 7-way line
+30,True,Architecture: Layers,"Architecture: 512/8/8 (Layers Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_30,2000,1000,1000,0.45619100000000007,0.45589222242929406,0.54380899999999999,0.51446986561882524,0.54380899999999999,0.51446986561882524,flipped,-1.7395205078125,0.55665361796016877,legacy 7-way line
+31,True,Architecture: Layers,"Architecture: 512/8/12 (Layers Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_31,2000,1000,1000,0.4259715,0.43679676089593727,0.57402850000000005,0.53994682513217307,0.57402850000000005,0.53994682513217307,flipped,-1.6108261718750001,0.60421080875154898,legacy 7-way line
+32,True,Learning Rate,"Learning Rate: LR 0.00015 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_32,2000,1000,1000,0.42451099999999997,0.43627663554400437,0.57548899999999992,0.53441744106541855,0.57548899999999992,0.53441744106541855,flipped,-1.9026877441406249,0.59221836909049941,legacy 7-way line
+33,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_33,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+34,True,Learning Rate,"Learning Rate: LR 0.00025 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_34,2000,1000,1000,0.44230199999999997,0.44792572699102756,0.55769800000000003,0.52204684601120832,0.55769800000000003,0.52204684601120832,flipped,-1.9074374999999999,0.51251080138198368,legacy 7-way line
+35,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.05, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_35,2000,1000,1000,0.43612050000000002,0.44149918528197146,0.56387949999999998,0.5249918601594864,0.56387949999999998,0.5249918601594864,flipped,-1.7743427734374999,0.49988898253195618,legacy 7-way line
+36,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.15, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_36,2000,1000,1000,0.44933299999999998,0.44507672141071813,0.55066700000000002,0.51606515195057334,0.55066700000000002,0.51606515195057334,flipped,-1.991265625,0.54427927926560882,legacy 7-way line
+37,True,Learning Rate,"Learning Rate: LR 0.0002 (polynomial, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_37,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+38,True,Window Size,"Window Size: 64bp (MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_38,2000,1000,1000,0.43197300000000005,0.44235561625600162,0.56802699999999995,0.53583940331881441,0.56802699999999995,0.53583940331881441,flipped,-1.7001718750000001,0.62569619384310315,legacy window sweep
+39,True,Window Size,"Window Size: 256bp (MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_39,2000,1000,1000,0.44268849999999998,0.45006563449986631,0.55731149999999996,0.53077842840062872,0.55731149999999996,0.53077842840062872,flipped,-1.7581943359375001,0.59361823386205825,legacy window sweep
+40,True,Learning Rate: Cosine,"Exp 06配置 + cosine LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_40,2000,1000,1000,0.44264550000000003,0.45551564996424543,0.55735449999999997,0.53668546691156038,0.55735449999999997,0.53668546691156038,flipped,-1.73626708984375,0.53929031343210965,legacy 7-way line
+41,True,Learning Rate: Cosine,"Exp 19配置 + cosine LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_41,2000,1000,1000,0.3987675,0.4204349370084246,0.60123250000000006,0.55804770868623743,0.60123250000000006,0.55804770868623743,flipped,-1.9218222656249999,0.55430695115805861,legacy 7-way line
+42,True,Learning Rate,"Exp 06配置 + polynomial LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_42,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+43,True,Learning Rate,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_43,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+44,True,Learning Rate,"Exp 06配置 + constant_with_warmup LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_44,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+45,True,Learning Rate,"Exp 06配置 + cosine LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_45,2000,1000,1000,0.44699649999999996,0.44988841993749179,0.55300349999999998,0.5191515015000302,0.55300349999999998,0.5191515015000302,flipped,-1.8558359375,0.57949372674740796,legacy 7-way line
+46,True,Learning Rate,"Exp 06配置 + polynomial LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_46,2000,1000,1000,0.44699649999999996,0.44988841993749179,0.55300349999999998,0.5191515015000302,0.55300349999999998,0.5191515015000302,flipped,-1.8558359375,0.57949372674740796,legacy 7-way line
+47,True,Learning Rate,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_47,2000,1000,1000,0.44699649999999996,0.44988841993749179,0.55300349999999998,0.5191515015000302,0.55300349999999998,0.5191515015000302,flipped,-1.8558359375,0.57949372674740796,legacy 7-way line
+48,True,Learning Rate,"Exp 19配置 + polynomial LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_48,2000,1000,1000,0.40176299999999998,0.43012193634461027,0.59823699999999991,0.57767086572283621,0.59823699999999991,0.57767086572283621,flipped,-1.7092465820312499,0.55927901688685333,legacy 7-way line
+49,True,Learning Rate,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_49,2000,1000,1000,0.40176299999999998,0.43012193634461027,0.59823699999999991,0.57767086572283621,0.59823699999999991,0.57767086572283621,flipped,-1.7092465820312499,0.55927901688685333,legacy 7-way line
+50,True,Learning Rate,"Exp 19配置 + constant_with_warmup LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_50,2000,1000,1000,0.43541449999999998,0.4527163393676894,0.56458549999999996,0.53379524412504831,0.56458549999999996,0.53379524412504831,flipped,-1.762480712890625,0.61734542629016287,legacy 7-way line
+51,True,Learning Rate,"Exp 19配置 + cosine LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_51,2000,1000,1000,0.41094950000000002,0.4248567083663824,0.58905049999999992,0.54195552349265463,0.58905049999999992,0.54195552349265463,flipped,-1.6808125,0.50088826518418128,legacy 7-way line
+52,True,Learning Rate,"Exp 19配置 + polynomial LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_52,2000,1000,1000,0.41094950000000002,0.4248567083663824,0.58905049999999992,0.54195552349265463,0.58905049999999992,0.54195552349265463,flipped,-1.6808125,0.50088826518418128,legacy 7-way line
+53,True,Learning Rate,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_53,2000,1000,1000,0.41094950000000002,0.4248567083663824,0.58905049999999992,0.54195552349265463,0.58905049999999992,0.54195552349265463,flipped,-1.6808125,0.50088826518418128,legacy 7-way line
+54,True,Ensemble,"Exp 06配置 + seed=42 (Ensemble member 1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_54,2000,1000,1000,0.42496150000000005,0.43872665749599238,0.57503850000000001,0.54300286080901561,0.57503850000000001,0.54300286080901561,flipped,-1.9408955078125001,0.59519239478325259,legacy 7-way line
+55,True,Ensemble,"Exp 06配置 + seed=123 (Ensemble member 2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_55,2000,1000,1000,0.41656149999999992,0.43540659496073786,0.58343849999999997,0.56346354705286328,0.58343849999999997,0.56346354705286328,flipped,-1.805287109375,0.65291470660524253,legacy 7-way line
+56,True,Ensemble,"Exp 06配置 + seed=456 (Ensemble member 3, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_56,2000,1000,1000,0.40185899999999997,0.42693961029800187,0.59814100000000003,0.56564582307705569,0.59814100000000003,0.56564582307705569,flipped,-1.6161582031249999,0.53146009270249583,legacy 7-way line
+57,True,Priority A,"Architecture sweep: 768/8/6 (Exp6 hyperparams, MLM 0.1, Soft Mask 0.1, LR 0.0002, 128bp)",full_training_57,2000,1000,1000,0.47745000000000004,0.46900574437846343,0.52254999999999996,0.51742174054398837,0.52254999999999996,0.51742174054398837,flipped,-1.3318505859375001,0.48430865986505606,legacy 7-way line
+58,True,Priority A,"Architecture sweep: 512/8/8 (Exp6 hyperparams, MLM 0.1, Soft Mask 0.1, LR 0.0002, 128bp)",full_training_58,2000,1000,1000,0.45619100000000007,0.45589222242929406,0.54380899999999999,0.51446986561882524,0.54380899999999999,0.51446986561882524,flipped,-1.7395205078125,0.55665361796016877,legacy 7-way line
+59,True,Priority A,"Regularization: weight_decay=0.05 (Exp6配置, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_59,2000,1000,1000,0.44959149999999998,0.44739156830391225,0.55040850000000008,0.51269587951233442,0.55040850000000008,0.51269587951233442,flipped,-1.8232626953125,0.5257465427582495,legacy 7-way line
+60,True,Priority A,"Regularization: weight_decay=0.1 (Exp6配置, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_60,2000,1000,1000,0.42086950000000001,0.43884595202951521,0.57913049999999999,0.53625837920026509,0.57913049999999999,0.53625837920026509,flipped,-1.80266357421875,0.6847249487129331,legacy 7-way line
+61,True,Priority A: LR Scheduler,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_61,2000,1000,1000,0.42930049999999997,0.43799354678825009,0.57069949999999992,0.52908633913069236,0.57069949999999992,0.52908633913069236,flipped,-1.9465859375000001,0.50925933796369671,legacy 7-way line
+62,True,Priority A: LR Scheduler,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_62,2000,1000,1000,0.40176299999999998,0.43012193634461027,0.59823699999999991,0.57767086572283621,0.59823699999999991,0.57767086572283621,flipped,-1.7092465820312499,0.55927901688685333,legacy 7-way line
+65,True,Aux Features: All Species,"Exp 06 config + all 6 species as aux_features (MLM 0.1, LR 0.0002, 512/8/6, n_aux=30)",full_training_65,2000,1000,1000,0.38141049999999999,0.40925057502409229,0.61858950000000013,0.57229872881421484,0.61858950000000013,0.57229872881421484,flipped,-2.0189887695312501,0.69748279353405396,legacy 7-way line
+68,True,Paper Reproduction + Eval Frequency Optimization,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=500)",full_training_68,2000,1000,1000,0.35809600000000008,0.39861887074652735,0.64190400000000003,0.60222455206516368,0.64190400000000003,0.60222455206516368,flipped,-1.9627822265624999,0.77448299371002549,legacy 7-way line
+69,True,Paper Reproduction + Eval Frequency Optimization,"Paper hyperparams + n_aux=3 + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=3, eval_steps=500)",full_training_69,2000,1000,1000,0.41301200000000005,0.4353337856366441,0.58698799999999995,0.57101500632595426,0.58698799999999995,0.57101500632595426,flipped,-2.4062968749999998,0.68804576898968672,legacy 7-way line
+70_seed123,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=123",full_training_70_paper_correct_seed123,2000,1000,1000,0.38460999999999995,0.40821974511714637,0.61538999999999999,0.55913165556696864,0.61538999999999999,0.55913165556696864,flipped,-2.0326132812500002,0.66594407376607478,other
+70_seed42,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=42",full_training_70_paper_correct_seed42,2000,1000,1000,0.35534149999999998,0.39699384032333107,0.64465850000000002,0.58617634438169941,0.64465850000000002,0.58617634438169941,flipped,-1.9661367187500001,0.61539556964448605,other
+70_seed456,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=456",full_training_70_paper_correct_seed456,2000,1000,1000,0.39680900000000002,0.41564913660681457,0.60319100000000003,0.55480924208912863,0.60319100000000003,0.55480924208912863,flipped,-2.0683281249999999,0.72587952964161651,other
+71_seed123,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=123",full_training_71_optimized_seed123,2000,1000,1000,0.36684499999999998,0.40071229926373808,0.63315499999999991,0.57811372372825642,0.63315499999999991,0.57811372372825642,flipped,-2.0937187499999999,0.77711338389309792,other
+71_seed42,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=42",full_training_71_optimized_seed42,2000,1000,1000,0.38145950000000001,0.4081489962840702,0.61854050000000005,0.5635723537982732,0.61854050000000005,0.5635723537982732,flipped,-1.9536914062499999,0.65145744636783498,other
+71_seed456,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=456",full_training_71_optimized_seed456,2000,1000,1000,0.34803700000000004,0.39413223864861752,0.65196300000000007,0.60467122899739945,0.65196300000000007,0.60467122899739945,flipped,-2.19737890625,0.73367534572465654,other
+72,True,Window Size: 512bp,"Exp 71优化 + 512bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=9 (1 epoch), batch=2048, max_position_embeddings=512",full_training_72_512bp,2000,1000,1000,0.38508450000000005,0.40833929808214292,0.61491549999999995,0.55875930466513357,0.61491549999999995,0.55875930466513357,flipped,-2.0365136718749999,0.72350064417738535,legacy window sweep
+73,True,Window Size: 1024bp,"Exp 71优化 + 1024bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=4 (1 epoch), batch=2048, max_position_embeddings=1024",full_training_73_1024bp,2000,1000,1000,0.38834049999999998,0.41050274540448783,0.61165950000000002,0.56065388267872684,0.61165950000000002,0.56065388267872684,flipped,-2.0408945312500002,0.79550562141969006,legacy window sweep
+74,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_74_64bp,2000,1000,1000,0.3546455,0.39659747922649774,0.64535449999999994,0.60768467451508568,0.64535449999999994,0.60768467451508568,flipped,-2.2170859374999998,0.80780499652734927,legacy window sweep
+74_seed123,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64, seed=123",full_training_74_64bp_seed123,2000,1000,1000,0.3968005,0.41540422835317153,0.6031995,0.55932528241956814,0.6031995,0.55932528241956814,flipped,-2.0319531249999998,0.6887026457249279,legacy window sweep
+74_seed456,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64, seed=456",full_training_74_64bp_seed456,2000,1000,1000,0.43643999999999999,0.44158147135870424,0.56355999999999995,0.53234945718667814,0.56355999999999995,0.53234945718667814,flipped,-2.0478598632812499,1.1830198447092881,legacy window sweep
+75,True,Window Size: 256bp,"Exp 71优化 + 256bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=18 (1 epoch), batch=2048, max_position_embeddings=256",full_training_75_256bp,2000,1000,1000,0.40219300000000002,0.41850593657512192,0.59780699999999998,0.54976337401768549,0.59780699999999998,0.54976337401768549,flipped,-1.9704238281249999,0.7059489610907651,legacy window sweep
+76,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + LR=0.00015: MLM 0.1, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_76_64bp_LR_0.00015,2000,1000,1000,0.39856449999999999,0.41551251688795998,0.60143550000000001,0.54741189655675848,0.60143550000000001,0.54741189655675848,flipped,-2.0356601562500001,0.69453264903147549,legacy 7-way line
+77,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + LR=0.00025: MLM 0.1, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_77_64bp_LR_0.00025,2000,1000,1000,0.38541249999999999,0.40828035036434818,0.61458750000000006,0.56006477228388962,0.61458750000000006,0.56006477228388962,flipped,-2.2823046874999999,0.76912547895669869,legacy 7-way line
+78,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + warmup=500: MLM 0.1, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_78_64bp_warmup_500,2000,1000,1000,0.40324650000000006,0.42324144058748098,0.59675349999999994,0.56029548884524438,0.59675349999999994,0.56029548884524438,flipped,-1.918888671875,0.72874763857162816,legacy 7-way line
+79,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + warmup=2000: MLM 0.1, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_79_64bp_warmup_2000,2000,1000,1000,0.37298550000000003,0.40550152035997677,0.62701450000000003,0.57753795018086285,0.62701450000000003,0.57753795018086285,flipped,-1.8960390625000001,0.71014162831310179,legacy 7-way line
+80,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.08: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_80_64bp_MLM_0.08,2000,1000,1000,0.43051699999999998,0.44150531436118678,0.56948300000000018,0.54511997732470729,0.56948300000000018,0.54511997732470729,flipped,-1.8644238281250001,1.0153669890402726,legacy 7-way line
+81,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.12: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_81_64bp_MLM_0.12,2000,1000,1000,0.39282949999999994,0.41482555380282404,0.60717050000000006,0.55852818591688558,0.60717050000000006,0.55852818591688558,flipped,-1.92664453125,0.66544042655105662,legacy 7-way line
+82,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + MLM=0.12: cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_82_64bp_LR_0.00025_MLM_0.12,2000,1000,1000,0.384824,0.41686658302877083,0.61517599999999995,0.5765070058611772,0.61517599999999995,0.5765070058611772,flipped,-2.2793789062499998,1.0561543298062694,legacy 7-way line
+83,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + warmup=500: MLM 0.1, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_83_64bp_LR_0.00025_warmup_500,2000,1000,1000,0.39784150000000001,0.41797328091068853,0.60215850000000004,0.55814484114788654,0.60215850000000004,0.55814484114788654,flipped,-1.847177734375,0.716536593727923,legacy 7-way line
+84,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + warmup=2000: MLM 0.1, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_84_64bp_LR_0.00025_warmup_2000,2000,1000,1000,0.40372099999999994,0.42064187598852854,0.596279,0.54802726720413519,0.596279,0.54802726720413519,flipped,-1.996142578125,0.80425384006170453,legacy 7-way line
+87,True,Hyperparameter Tuning: 64bp,"Exp 82优化 + 64bp窗口 + LR=0.00025 + MLM=0.12 + warmup=500: cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_87_64bp_LR_0.00025_MLM_0.12_warmup_500,2000,1000,1000,0.37548500000000001,0.40483248751526363,0.62451499999999993,0.57215291454084272,0.62451499999999993,0.57215291454084272,flipped,-2.3629101562499999,1.0280857881529752,legacy 7-way line
+88,True,Hyperparameter Tuning: 64bp,"Exp 82优化 + 64bp窗口 + LR=0.00025 + MLM=0.12 + warmup=2000: cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_88_64bp_LR_0.00025_MLM_0.12_warmup_2000,2000,1000,1000,0.37817250000000008,0.40604749463555984,0.62182749999999998,0.57259364034234772,0.62182749999999998,0.57259364034234772,flipped,-1.94583984375,0.69108731802651746,legacy 7-way line
+92,True,Multi-Seed Validation: 64bp,"Exp 80多种子验证 + 64bp窗口 + seed=123: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_92_64bp_MLM_0.08_seed123,2000,1000,1000,0.42703400000000002,0.44074379741177105,0.57296599999999998,0.55070622715148931,0.57296599999999998,0.55070622715148931,flipped,-2.0693125000000001,1.1897953791259459,legacy 7-way line
+93,True,Multi-Seed Validation: 64bp,"Exp 80多种子验证 + 64bp窗口 + seed=456: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_93_64bp_MLM_0.08_seed456,2000,1000,1000,0.37075349999999996,0.40525475094547236,0.62924650000000004,0.59770545727801261,0.62924650000000004,0.59770545727801261,flipped,-2.2130097656249998,1.0270126239015602,legacy 7-way line
+94,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + LR=0.00025: cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_94_64bp_MLM_0.08_LR_0.00025,2000,1000,1000,0.37986350000000002,0.40804746981769635,0.62013649999999998,0.56822012096805141,0.62013649999999998,0.56822012096805141,flipped,-1.9615605468749999,0.7084740981762937,legacy 7-way line
+95,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + warmup=500: LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_95_64bp_MLM_0.08_warmup_500,2000,1000,1000,0.40545100000000001,0.42230839704630363,0.59454899999999999,0.55322178302827973,0.59454899999999999,0.55322178302827973,flipped,-1.847392578125,0.7193618207796435,legacy 7-way line
+96,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + warmup=2000: LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_96_64bp_MLM_0.08_warmup_2000,2000,1000,1000,0.38562749999999996,0.40890739859350356,0.61437249999999999,0.55411922435636407,0.61437249999999999,0.55411922435636407,flipped,-2.1481953125,0.74470027097571856,legacy 7-way line
+97,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.06: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_97_64bp_MLM_0.06,2000,1000,1000,0.39233200000000001,0.4131847641300157,0.60766800000000021,0.55347863529780028,0.60766800000000021,0.55347863529780028,flipped,-1.80104296875,0.65605028014437627,legacy 7-way line
+98,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.09: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_98_64bp_MLM_0.09,2000,1000,1000,0.38091500000000006,0.40730768158356417,0.61908499999999989,0.57470219434511172,0.61908499999999989,0.57470219434511172,flipped,-2.0515234375000002,0.73209983440887616,legacy 7-way line
+99,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.07: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_99_64bp_MLM_0.07,2000,1000,1000,0.39220349999999993,0.41438111195322391,0.60779650000000007,0.56576604782874496,0.60779650000000007,0.56576604782874496,flipped,-1.9748281249999999,0.70738185005491161,legacy 7-way line
+,False,External,GPN-MSA trained checkpoint (not listed in base-model ablation registry),gpn_msa_trained,2000,1000,1000,0.30716700000000002,0.37667541452960018,0.69283300000000003,0.6579185849419642,0.69283300000000003,0.6579185849419642,flipped,-3.0486369156390429,1.8615944903139912,gpn-style checkpoint

thesis_repro/phylo_msa1_outputs/VEP_ALL_BASE_MODELS_STRICTV2_MATCHED_METRICS_CI.csv

@@ -0,0 +1,176 @@
+model,model_display,n,pos,neg,polarity,auroc_cal,auroc_ci_low,auroc_ci_high,auprc_cal,auprc_ci_low,auprc_ci_high
+full_training_207,Exp. 207,2000,1000,1000,flipped,0.733943,0.7125272947801506,0.755020371391086,0.6847323524773955,0.6533980458282247,0.7186731616314354
+full_training_210,Exp. 210,2000,1000,1000,flipped,0.7262250000000001,0.7050979101994459,0.7468784719062471,0.6740753518571917,0.6423441740438786,0.7094963575058489
+full_training_206,Exp. 206,2000,1000,1000,flipped,0.725195,0.7032836364229388,0.7468506849830528,0.6691809398692765,0.6370310657718485,0.7031172905021881
+full_training_169,Exp. 169,2000,1000,1000,flipped,0.7239895000000001,0.7006639321766578,0.7444342453138575,0.6799283051027143,0.6489149925234776,0.7119381023560871
+full_training_209,Exp. 209,2000,1000,1000,flipped,0.718224,0.6962811138204397,0.7393275083430835,0.6662996670808599,0.6349055972837654,0.6999146079409072
+gpn_msa_trained,gpn\_msa\_trained,2000,1000,1000,flipped,0.692833,0.6699511182402567,0.7155528619662951,0.6579185849419642,0.627168339823677,0.6912906186345171
+full_training_201,Exp. 201,2000,1000,1000,flipped,0.6738085,0.650918558597175,0.6972209171517794,0.6293574033355777,0.5973498327311941,0.6633426161868698
+full_training_202,Exp. 202,2000,1000,1000,flipped,0.6654845,0.643022224664312,0.6887464733019331,0.624581599809271,0.5937271660108828,0.6591000393519586
+full_training_204,Exp. 204,2000,1000,1000,flipped,0.6625669999999999,0.6395363871218602,0.6864060297487499,0.6216540233650435,0.5907964290109966,0.6573355313579569
+full_training_203,Exp. 203,2000,1000,1000,flipped,0.6567280000000001,0.6338650048914087,0.6810401717928908,0.6173986577145278,0.5870013020276517,0.6521139293163132
+full_training_71_optimized_seed456,Exp. 71,2000,1000,1000,flipped,0.6519630000000001,0.6285995182116811,0.6752073919477776,0.6046712289973994,0.5736162644702323,0.638133640701809
+full_training_74_64bp,Exp. 74,2000,1000,1000,flipped,0.6453544999999999,0.6224719029412621,0.669327657617609,0.6076846745150857,0.5759849145623456,0.6416891469569679
+full_training_70_paper_correct_seed42,Exp. 70,2000,1000,1000,flipped,0.6446585,0.6203051820888035,0.6694234686216052,0.5861763443816994,0.5557085394437483,0.6204051964954022
+full_training_68,Exp. 68,2000,1000,1000,flipped,0.641904,0.6177526678257178,0.6655567009072226,0.6022245520651637,0.5705969110018535,0.6368951947953213
+full_training_140,Exp. 140,2000,1000,1000,flipped,0.641006,0.6171827660314203,0.6661387617933114,0.5851556156659712,0.5556301318063549,0.6187524852662529
+full_training_116,Exp. 116,2000,1000,1000,flipped,0.636169,0.6128357360238853,0.6608071385443519,0.5907784603434083,0.5600512983451231,0.6247404818724124
+full_training_136,Exp. 136,2000,1000,1000,flipped,0.6357485,0.611666580064641,0.6612842821229855,0.5826107903212173,0.5526300636726031,0.6162286261808426
+full_training_128,Exp. 128,2000,1000,1000,flipped,0.6335504999999999,0.6097407221357397,0.6573760224578192,0.5930540800385531,0.5616419546699631,0.6268414138200935
+full_training_71_optimized_seed123,Exp. 71,2000,1000,1000,flipped,0.6331549999999999,0.6088744439629217,0.6569131352179608,0.5781137237282564,0.5482468797157943,0.6132660301093995
+full_training_122,Exp. 122,2000,1000,1000,flipped,0.6296935,0.6055507014131059,0.6531998426418304,0.5978725980041113,0.5667961662305656,0.6321613370865639
+full_training_93_64bp_MLM_0.08_seed456,Exp. 93,2000,1000,1000,flipped,0.6292465,0.6055788449662458,0.6534338298222564,0.5977054572780126,0.5660459497756221,0.6305629775580249
+full_training_132,Exp. 132,2000,1000,1000,flipped,0.627602,0.6038268056833523,0.652259639697567,0.5629790256682738,0.5326361728512622,0.595893605486096
+full_training_79_64bp_warmup_2000,Exp. 79,2000,1000,1000,flipped,0.6270145,0.6031916183041406,0.6511535001086851,0.5775379501808628,0.5460881617080454,0.6120737214046809
+full_training_87_64bp_LR_0.00025_MLM_0.12_warmup_500,Exp. 87,2000,1000,1000,flipped,0.6245149999999999,0.6001623007574748,0.6494939892560161,0.5721529145408427,0.5415599755829553,0.605661747299967
+full_training_108_64bp_focal_loss,Exp. 108,2000,1000,1000,flipped,0.623903,0.5996971269611818,0.6480159863833064,0.573526024515667,0.5429140769254933,0.6075402150239689
+full_training_111_64bp_focal_loss_tuned,Exp. 111,2000,1000,1000,flipped,0.6235395,0.6002707310716459,0.6469409942678794,0.563950491321207,0.534853606222056,0.5970775949818166
+full_training_162,Exp. 162,2000,1000,1000,flipped,0.62216,0.5974750251480958,0.6460982884018527,0.5689456013571603,0.5383657357512796,0.6027417304264808
+full_training_88_64bp_LR_0.00025_MLM_0.12_warmup_2000,Exp. 88,2000,1000,1000,flipped,0.6218275,0.5975037000677675,0.6461605494551717,0.5725936403423477,0.5429478963530464,0.6051897109752684
+full_training_141,Exp. 141,2000,1000,1000,flipped,0.621526,0.5978353200975132,0.646026886474468,0.5874731409299327,0.5554224474715381,0.6219443603111879
+full_training_157,Exp. 157,2000,1000,1000,flipped,0.620168,0.5951968646321489,0.645739709973441,0.5741515957934253,0.5447599483605309,0.6083881418390938
+full_training_94_64bp_MLM_0.08_LR_0.00025,Exp. 94,2000,1000,1000,flipped,0.6201365,0.5959231304796425,0.6439012052590893,0.5682201209680514,0.5378550185348837,0.6010691788643644
+full_training_125,Exp. 125,2000,1000,1000,flipped,0.6199135,0.5944759285195459,0.6449286573114607,0.5647577553612588,0.5337648970173652,0.5981233784769332
+full_training_129,Exp. 129,2000,1000,1000,flipped,0.619468,0.5945282836856647,0.6436940804326674,0.5807427871440674,0.5487810035983508,0.614113278368347
+full_training_101_64bp_LR_0.00018,Exp. 101,2000,1000,1000,flipped,0.6193030000000002,0.5956983218307289,0.6434722314702462,0.567958909988223,0.5383106714350786,0.6021053749435528
+full_training_98_64bp_MLM_0.09,Exp. 98,2000,1000,1000,flipped,0.6190849999999999,0.5957854617860722,0.6426469981659845,0.5747021943451117,0.5449650878044662,0.6090197086174116
+full_training_139,Exp. 139,2000,1000,1000,flipped,0.6186885,0.5938982114728443,0.6423538763172392,0.5787542315944223,0.5490372048323651,0.6108933422817264
+full_training_126,Exp. 126,2000,1000,1000,flipped,0.6186594999999999,0.5937146043120999,0.642589060474243,0.5665183401758486,0.5370781492289881,0.5988830226620359
+full_training_65,Exp. 65,2000,1000,1000,flipped,0.6185895000000001,0.5928271478966965,0.642308647510589,0.5722987288142148,0.5426529730519905,0.6064657977285896
+full_training_71_optimized_seed42,Exp. 71,2000,1000,1000,flipped,0.6185405,0.5943485522996611,0.6437545329791655,0.5635723537982732,0.534307777024282,0.5973050440977941
+full_training_119,Exp. 119,2000,1000,1000,flipped,0.6183285,0.594007078745685,0.642910653966292,0.5702132207151084,0.5402274345553993,0.6038035544489703
+full_training_112_64bp_combined_loss,Exp. 112,2000,1000,1000,flipped,0.6179330000000001,0.59291618058236,0.6418285686997588,0.5567226997158773,0.5290487981159859,0.5887140647861071
+full_training_103_64bp_warmup_750,Exp. 103,2000,1000,1000,flipped,0.6176425,0.5943196562157096,0.6413563305947754,0.5548445873767158,0.5272496784319965,0.5871606633499618
+full_training_127,Exp. 127,2000,1000,1000,flipped,0.617261,0.5926486706319901,0.6409053545746987,0.5644035202941005,0.5332174845371269,0.596899907104737
+full_training_131,Exp. 131,2000,1000,1000,flipped,0.615785,0.5903984873981942,0.6398515765646116,0.5616351398558378,0.5310137615998193,0.594727977871724
+full_training_105_64bp_warmup_1500,Exp. 105,2000,1000,1000,flipped,0.61578,0.5916766190040952,0.639461071576052,0.551904075223377,0.5242463894068995,0.5842701414077281
+full_training_70_paper_correct_seed123,Exp. 70,2000,1000,1000,flipped,0.61539,0.5909018188736561,0.6409417058678699,0.5591316555669686,0.5300517235900335,0.5923042201203771
+full_training_137,Exp. 137,2000,1000,1000,flipped,0.615374,0.5905518994739826,0.6390414414358062,0.5721953645006291,0.5430131003966918,0.6059103225008469
+full_training_82_64bp_LR_0.00025_MLM_0.12,Exp. 82,2000,1000,1000,flipped,0.615176,0.5895835695284621,0.6388085047048796,0.5765070058611772,0.5441256426450501,0.6102113902690629
+full_training_124,Exp. 124,2000,1000,1000,flipped,0.6151519999999999,0.5907219768333591,0.6394059288688801,0.561569602602702,0.5302203120881016,0.5953630097924912
+full_training_72_512bp,Exp. 72,2000,1000,1000,flipped,0.6149155,0.590664571486667,0.6386160904180906,0.5587593046651336,0.5296156058804635,0.5918980849755717
+full_training_77_64bp_LR_0.00025,Exp. 77,2000,1000,1000,flipped,0.6145875000000001,0.5905500760949035,0.6395499097463013,0.5600647722838896,0.5313972598391423,0.5937714526496369
+full_training_96_64bp_MLM_0.08_warmup_2000,Exp. 96,2000,1000,1000,flipped,0.6143725,0.5903434123952236,0.6386055444734287,0.5541192243563641,0.52606416794269,0.5860940116401511
+full_training_113_64bp_focal_loss_gamma25,Exp. 113,2000,1000,1000,flipped,0.6138855,0.5894960126353981,0.6375497388338417,0.5765083916981948,0.5466491343850968,0.6096755789715934
+full_training_107_64bp_class_weights,Exp. 107,2000,1000,1000,flipped,0.6133185,0.5877264986864731,0.6365192252781537,0.5760443050678346,0.5443569488924975,0.6103251858139401
+full_training_102_64bp_LR_0.00022,Exp. 102,2000,1000,1000,flipped,0.6129395,0.5886504279905067,0.6363868584362242,0.5615053894942705,0.5325523350465109,0.595192640703998
+full_training_118,Exp. 118,2000,1000,1000,flipped,0.6120765,0.5877035249106464,0.6374247772784838,0.5533321231816615,0.5237746579526972,0.5855208114109368
+full_training_134,Exp. 134,2000,1000,1000,flipped,0.6117319999999999,0.5873615834354773,0.6359813235741454,0.559497353410816,0.5287590523472613,0.593010688683243
+full_training_73_1024bp,Exp. 73,2000,1000,1000,flipped,0.6116595,0.5869651535017218,0.6365305905751167,0.5606538826787268,0.5294835136675561,0.5949720466461732
+full_training_110_64bp_class_weights_aggressive,Exp. 110,2000,1000,1000,flipped,0.6109864999999999,0.5865182877159214,0.6356317856548124,0.5678344446090657,0.5379875102966003,0.6002530607409449
+full_training_123,Exp. 123,2000,1000,1000,flipped,0.6099870000000001,0.5850943815717474,0.6342037569056059,0.5550381964669231,0.525863004237781,0.5878095772229691
+full_training_120,Exp. 120,2000,1000,1000,flipped,0.6094109999999999,0.5852703087555154,0.6343973930123453,0.5556453769853106,0.5261898640458166,0.587978637663742
+full_training_138,Exp. 138,2000,1000,1000,flipped,0.6091179999999999,0.5839940766481134,0.633789265311381,0.5786432228126065,0.5475685320549212,0.6123343398097352
+full_training_133,Exp. 133,2000,1000,1000,flipped,0.6078975,0.5833028892505969,0.6323037921890279,0.5539978118192095,0.5236683558394757,0.5866144655741935
+full_training_99_64bp_MLM_0.07,Exp. 99,2000,1000,1000,flipped,0.6077965000000001,0.5838974424338638,0.6315563407828391,0.565766047828745,0.5358962338276307,0.5988470684139918
+full_training_97_64bp_MLM_0.06,Exp. 97,2000,1000,1000,flipped,0.6076680000000002,0.5835292234078628,0.6332338121155391,0.5534786352978003,0.5241455721875509,0.5866469071069091
+full_training_81_64bp_MLM_0.12,Exp. 81,2000,1000,1000,flipped,0.6071705000000001,0.584190616427527,0.6315225551526674,0.5585281859168856,0.5291008182760877,0.5905147713349971
+full_training_165,Exp. 165,2000,1000,1000,flipped,0.6069694999999999,0.5817252385812437,0.6309947039912044,0.5566995809788468,0.5279763405097518,0.5893059958660923
+full_training_144,Exp. 144,2000,1000,1000,flipped,0.6069694999999999,0.5817252385812437,0.6309947039912044,0.5566995809788468,0.5279763405097518,0.5893059958660923
+full_training_142,Exp. 142,2000,1000,1000,flipped,0.6068724999999999,0.5825870894809085,0.6326522188689483,0.5519408466138951,0.5240774111045878,0.5843166660221615
+full_training_135,Exp. 135,2000,1000,1000,flipped,0.6066775,0.5816127343045084,0.6305611160809282,0.5570269347343862,0.5280506465617183,0.5902886527131415
+full_training_130,Exp. 130,2000,1000,1000,flipped,0.6056015,0.5816497107956974,0.6299816014788481,0.5538971696411383,0.5252992251130204,0.5864912365560482
+full_training_106_64bp_soft_mask_0.15,Exp. 106,2000,1000,1000,flipped,0.605224,0.5808306119996166,0.628468226218332,0.5573527895613444,0.5294061461004819,0.589704785673158
+full_training_74_64bp_seed123,Exp. 74,2000,1000,1000,flipped,0.6031995,0.5789646346826413,0.6290102170340831,0.5593252824195681,0.5297720076141177,0.5929364454110664
+full_training_70_paper_correct_seed456,Exp. 70,2000,1000,1000,flipped,0.603191,0.5783193532195801,0.6289758473062173,0.5548092420891286,0.5253694117016099,0.5885096631974441
+full_training_115,Exp. 115,2000,1000,1000,flipped,0.6027935,0.5778465048531563,0.6285264234175083,0.5472860643328314,0.5183487338606332,0.5798254637230171
+full_training_104_64bp_warmup_1250,Exp. 104,2000,1000,1000,flipped,0.6025415,0.5776830190088728,0.6268215122569987,0.5591123557626467,0.5288234432980482,0.5933376932011096
+full_training_83_64bp_LR_0.00025_warmup_500,Exp. 83,2000,1000,1000,flipped,0.6021585,0.5770224385952579,0.6273372188061351,0.5581448411478865,0.5274881501920617,0.5923610597746296
+full_training_76_64bp_LR_0.00015,Exp. 76,2000,1000,1000,flipped,0.6014355,0.5777480928219966,0.6257962468997555,0.5474118965567585,0.5183368822147453,0.5783768759939864
+full_training_41,Exp. 41,2000,1000,1000,flipped,0.6012325000000001,0.5763799532695333,0.6273561947398272,0.5580477086862374,0.5282745484298793,0.5921584753693837
+full_training_117,Exp. 117,2000,1000,1000,flipped,0.601189,0.5763510211147039,0.6263368992851086,0.5443890596325135,0.5156260350710011,0.5765613954181554
+full_training_48,Exp. 48,2000,1000,1000,flipped,0.5982369999999999,0.5732654608517692,0.6224666238850585,0.5776708657228362,0.5471978355212018,0.6092369405448976
+full_training_62,Exp. 62,2000,1000,1000,flipped,0.5982369999999999,0.5732654608517692,0.6224666238850585,0.5776708657228362,0.5471978355212018,0.6092369405448976
+full_training_49,Exp. 49,2000,1000,1000,flipped,0.5982369999999999,0.5732654608517692,0.6224666238850585,0.5776708657228362,0.5471978355212018,0.6092369405448976
+full_training_56,Exp. 56,2000,1000,1000,flipped,0.598141,0.5731630216031812,0.6225623735377178,0.5656458230770557,0.5345640714838866,0.5986374130852815
+full_training_75_256bp,Exp. 75,2000,1000,1000,flipped,0.597807,0.5731363657127155,0.622403813113226,0.5497633740176855,0.5208713668077004,0.5819983951958413
+full_training_78_64bp_warmup_500,Exp. 78,2000,1000,1000,flipped,0.5967534999999999,0.5729892439279988,0.6208459387712546,0.5602954888452444,0.5311546882550656,0.5930638322199311
+full_training_84_64bp_LR_0.00025_warmup_2000,Exp. 84,2000,1000,1000,flipped,0.596279,0.5725784436978606,0.6211725404102777,0.5480272672041352,0.5180680854151354,0.5803703659708463
+full_training_121,Exp. 121,2000,1000,1000,flipped,0.5959220000000001,0.5719670036763129,0.619948159275998,0.5734722460870275,0.5428579768341258,0.6058931997368777
+full_training_95_64bp_MLM_0.08_warmup_500,Exp. 95,2000,1000,1000,flipped,0.594549,0.5694917322428065,0.620477574635641,0.5532217830282797,0.5237338695530905,0.5870478291309075
+full_training_20,Exp. 20,2000,1000,1000,flipped,0.5906279999999999,0.5668362544069454,0.6163403474539959,0.5489355152956636,0.5195218317696801,0.5818574763704201
+full_training_148,Exp. 148,2000,1000,1000,flipped,0.5894645000000001,0.5648140905068691,0.612979753766694,0.5647404620488217,0.5341848698531974,0.5978355135788638
+full_training_51,Exp. 51,2000,1000,1000,flipped,0.5890504999999999,0.5639455421070735,0.614585903391384,0.5419555234926546,0.512585412010482,0.5750113011892474
+full_training_52,Exp. 52,2000,1000,1000,flipped,0.5890504999999999,0.5639455421070735,0.614585903391384,0.5419555234926546,0.512585412010482,0.5750113011892474
+full_training_53,Exp. 53,2000,1000,1000,flipped,0.5890504999999999,0.5639455421070735,0.614585903391384,0.5419555234926546,0.512585412010482,0.5750113011892474
+full_training_21,Exp. 21,2000,1000,1000,flipped,0.5885419999999999,0.5648114480363134,0.6127596257564745,0.5653760134620599,0.53591826445844,0.5999006634002075
+full_training_69,Exp. 69,2000,1000,1000,flipped,0.586988,0.5614896858172627,0.6124851718259208,0.5710150063259543,0.5408671556967359,0.6044307846049608
+full_training_55,Exp. 55,2000,1000,1000,flipped,0.5834385,0.5579629335957123,0.6088697008015723,0.5634635470528633,0.5334031642194806,0.5960136111667935
+full_training_17,Exp. 17,2000,1000,1000,flipped,0.582562,0.5592278944130483,0.6077355794286101,0.5593841427327672,0.5294941341624342,0.5931971923054611
+full_training_15,Exp. 15,2000,1000,1000,flipped,0.581032,0.5566530459361083,0.6059215883632807,0.5370172317447502,0.5083568091487032,0.5688252507509154
+full_training_29,Exp. 29,2000,1000,1000,flipped,0.5801324999999999,0.5554762984956864,0.6042416934016724,0.5503745688085787,0.5212042287903904,0.5845986814088515
+full_training_23,Exp. 23,2000,1000,1000,flipped,0.5799350000000001,0.5547557090674439,0.6044808098205394,0.5597031216365916,0.5286135971280532,0.5915578574148477
+full_training_60,Exp. 60,2000,1000,1000,flipped,0.5791305,0.5542784924175049,0.6046111713527644,0.5362583792002651,0.5071445250039986,0.5678293029192689
+full_training_05,Exp. 5,2000,1000,1000,flipped,0.5787519999999999,0.5536154726246997,0.603542393853287,0.5436618498868525,0.5140172051416084,0.5773157476796213
+full_training_109_64bp_768_12_8,Exp. 109,2000,1000,1000,flipped,0.5782134999999999,0.5529067047291364,0.6032145619740876,0.5396438756274388,0.5111500899359269,0.5717635101204681
+full_training_08,Exp. 8,2000,1000,1000,flipped,0.5764825,0.5518826557586324,0.6009731825095121,0.5399104832284252,0.5104702193586241,0.5728318276921959
+full_training_26,Exp. 26,2000,1000,1000,flipped,0.5762235,0.5509505580522951,0.6005256270711468,0.5551072899901709,0.5261540861291614,0.5887552161627732
+full_training_155,Exp. 155,2000,1000,1000,flipped,0.575947,0.5513073667812214,0.6006763746894876,0.543601122720263,0.5147702041012139,0.5758398872363076
+full_training_24,Exp. 24,2000,1000,1000,flipped,0.5755215,0.5501809737607797,0.6003414807438425,0.5502261801831547,0.5208121608879345,0.582853517035392
+full_training_32,Exp. 32,2000,1000,1000,flipped,0.5754889999999999,0.5506591150852298,0.6009060933171886,0.5344174410654186,0.5054409600908468,0.5677574227837371
+full_training_54,Exp. 54,2000,1000,1000,flipped,0.5750385,0.5496069379526966,0.6000352008946799,0.5430028608090156,0.5129761950887832,0.5778386942947836
+full_training_31,Exp. 31,2000,1000,1000,flipped,0.5740285,0.5483952626879787,0.5988177398596024,0.5399468251321731,0.5101070990258533,0.5731692469859659
+full_training_92_64bp_MLM_0.08_seed123,Exp. 92,2000,1000,1000,flipped,0.572966,0.5474398765671317,0.5985744879781271,0.5507062271514893,0.5197025890369622,0.5843762846379801
+full_training_10,Exp. 10,2000,1000,1000,flipped,0.572678,0.5487912421036476,0.598045359289692,0.5308057933517494,0.5028822786531199,0.5615789134444653
+full_training_22,Exp. 22,2000,1000,1000,flipped,0.5726344999999999,0.5478138412415621,0.5978229370886969,0.5478864956412605,0.5179879904366504,0.5800305030361103
+full_training_143,Exp. 143,2000,1000,1000,flipped,0.5712775,0.5465512956012591,0.5957655726307988,0.5410140982554464,0.5103291225084409,0.5739941692102429
+full_training_167,Exp. 167,2000,1000,1000,flipped,0.5712775,0.5465512956012591,0.5957655726307988,0.5410140982554464,0.5103291225084409,0.5739941692102429
+full_training_61,Exp. 61,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_33,Exp. 33,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_43,Exp. 43,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_37,Exp. 37,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_42,Exp. 42,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_44,Exp. 44,2000,1000,1000,flipped,0.5706994999999999,0.5461190813176274,0.5947840514733507,0.5290863391306924,0.5008188431482685,0.5616442352612031
+full_training_25,Exp. 25,2000,1000,1000,flipped,0.570271,0.5445502789971505,0.5947329876691773,0.5423241304470768,0.5136212729108057,0.5744743528214029
+full_training_80_64bp_MLM_0.08,Exp. 80,2000,1000,1000,flipped,0.5694830000000002,0.5430685315742375,0.5946912234598885,0.5451199773247073,0.5154735651731103,0.5773840363501256
+full_training_114_64bp_MLM_0.08_retrain,Exp. 114,2000,1000,1000,flipped,0.5694830000000002,0.5430685315742375,0.5946912234598885,0.5451199773247073,0.5154735651731103,0.5773840363501256
+full_training_14,Exp. 14,2000,1000,1000,flipped,0.5688085,0.5444020909283058,0.5942026706566355,0.5312959844239147,0.5013203875649638,0.5636569198416915
+full_training_07,Exp. 7,2000,1000,1000,flipped,0.568254,0.5440786392829473,0.5927954036068426,0.5296417918916938,0.501651275440562,0.5618833202599036
+full_training_13,Exp. 13,2000,1000,1000,flipped,0.5681295000000001,0.5440761702610304,0.5928911584897375,0.5349908635659566,0.5065473737599215,0.5684945166889994
+full_training_38,Exp. 38,2000,1000,1000,flipped,0.568027,0.5433200565565283,0.5930095498990636,0.5358394033188144,0.5054936065360788,0.5680542783287861
+full_training_100_64bp_MLM_0.085,Exp. 100,2000,1000,1000,flipped,0.567925,0.5437960632925128,0.5932053116948538,0.5406918580202631,0.5098303799908541,0.5723011543395211
+full_training_06,Exp. 6,2000,1000,1000,flipped,0.5677559999999999,0.5420207385378304,0.5918958958695981,0.5485517438515886,0.5178598586914153,0.5821479864132837
+full_training_28,Exp. 28,2000,1000,1000,flipped,0.567466,0.5422926993531083,0.5930573222053478,0.5283892109193837,0.4988785989228046,0.559838774336631
+full_training_19,Exp. 19,2000,1000,1000,flipped,0.565863,0.5418490251457462,0.5907337390567194,0.5423058619459147,0.512365420492077,0.5754745696572751
+full_training_04,Exp. 4,2000,1000,1000,flipped,0.56551,0.5400533315130257,0.5900779748863498,0.525822873662503,0.49581954226137875,0.5567601618736047
+full_training_50,Exp. 50,2000,1000,1000,flipped,0.5645855,0.5402409809304379,0.5896544419459284,0.5337952441250483,0.5050677857397148,0.5666605450998895
+full_training_35,Exp. 35,2000,1000,1000,flipped,0.5638795,0.5381291518348106,0.5881456169026371,0.5249918601594864,0.49646103641050965,0.5562037018988667
+full_training_74_64bp_seed456,Exp. 74,2000,1000,1000,flipped,0.56356,0.536999096173286,0.5890986540524176,0.5323494571866781,0.5030640008881967,0.5641108948447766
+full_training_09,Exp. 9,2000,1000,1000,flipped,0.5626485000000001,0.5381076301937827,0.5874725668059513,0.5301481865676793,0.5033562323799238,0.5610261682224665
+full_training_02,Exp. 2,2000,1000,1000,flipped,0.5607075,0.5364807619365671,0.5855287865385462,0.5317118669641843,0.5027410058405986,0.5637718462878532
+full_training_161,Exp. 161,2000,1000,1000,flipped,0.5598865000000001,0.534403342301109,0.5852878574888664,0.5305850019248513,0.5010463920811676,0.5618342404895397
+full_training_34,Exp. 34,2000,1000,1000,flipped,0.557698,0.5327003217630656,0.5828758300265857,0.5220468460112083,0.4937757676145812,0.5544819546787705
+full_training_27,Exp. 27,2000,1000,1000,flipped,0.557674,0.5325658875723497,0.58245524944963,0.528488016864884,0.5001295539430051,0.5605458266941298
+full_training_40,Exp. 40,2000,1000,1000,flipped,0.5573545,0.5314632186371141,0.582222787790448,0.5366854669115604,0.5068168399690592,0.5704425169444862
+full_training_39,Exp. 39,2000,1000,1000,flipped,0.5573115,0.5325561326301735,0.5827256040843269,0.5307784284006287,0.5008894290693378,0.56421232701239
+full_training_16,Exp. 16,2000,1000,1000,flipped,0.5571605000000001,0.5323244458978136,0.582485956205123,0.520610040650009,0.49177513269862894,0.5515209727122689
+full_training_11,Exp. 11,2000,1000,1000,flipped,0.5568355,0.5319794212556531,0.5824956428654094,0.5436893392797401,0.513249550915882,0.5770230621637733
+full_training_18,Exp. 18,2000,1000,1000,flipped,0.5551715,0.5302772156158638,0.5792517090233869,0.5320873295181459,0.5022554228281626,0.5649189549582996
+full_training_150,Exp. 150,2000,1000,1000,flipped,0.5544600000000001,0.5291326275571454,0.5780645425514984,0.535069092409769,0.5041698057971894,0.5664156804282166
+full_training_166,Exp. 166,2000,1000,1000,flipped,0.5542140000000001,0.5288264128195862,0.5797696523140748,0.5277880480884656,0.49842139678051645,0.5597365771351864
+full_training_147,Exp. 147,2000,1000,1000,flipped,0.5531295,0.5270863728285532,0.5774824870636509,0.5500807070275537,0.5183683743055771,0.5831750914727568
+full_training_46,Exp. 46,2000,1000,1000,flipped,0.5530035,0.5279153264450765,0.5787069163789653,0.5191515015000302,0.491717238382704,0.5513585155319379
+full_training_45,Exp. 45,2000,1000,1000,flipped,0.5530035,0.5279153264450765,0.5787069163789653,0.5191515015000302,0.491717238382704,0.5513585155319379
+full_training_47,Exp. 47,2000,1000,1000,flipped,0.5530035,0.5279153264450765,0.5787069163789653,0.5191515015000302,0.491717238382704,0.5513585155319379
+full_training_01,Exp. 1,2000,1000,1000,flipped,0.5514595,0.5273761158460984,0.5760191891528326,0.5291494730905917,0.5005260382301675,0.5614491339645818
+full_training_36,Exp. 36,2000,1000,1000,flipped,0.550667,0.5254683279845079,0.5748707148478155,0.5160651519505733,0.4883738513293734,0.5469717030940165
+full_training_59,Exp. 59,2000,1000,1000,flipped,0.5504085000000001,0.5251538521069771,0.5754267434529171,0.5126958795123344,0.4852240147469222,0.5454504121717612
+full_training_12,Exp. 12,2000,1000,1000,flipped,0.544064,0.5201187569077387,0.5690352125976593,0.518489930430114,0.49151192948912403,0.5488676447977888
+full_training_58,Exp. 58,2000,1000,1000,flipped,0.543809,0.51887088249054,0.5682848234578346,0.5144698656188252,0.48623915756678965,0.5461227304178229
+full_training_30,Exp. 30,2000,1000,1000,flipped,0.543809,0.51887088249054,0.5682848234578346,0.5144698656188252,0.48623915756678965,0.5461227304178229
+full_training_154,Exp. 154,2000,1000,1000,flipped,0.5433074999999999,0.5166115096017939,0.5687908337223432,0.5358045101691986,0.5059961859966406,0.5680276576744586
+full_training_146,Exp. 146,2000,1000,1000,flipped,0.5433074999999999,0.5166115096017939,0.5687908337223432,0.5358045101691986,0.5059961859966406,0.5680276576744586
+full_training_159,Exp. 159,2000,1000,1000,flipped,0.5394125000000001,0.514045079928911,0.564851067040085,0.5279883514327945,0.4982607868403603,0.5606491470630786
+full_training_168,Exp. 168,2000,1000,1000,flipped,0.536927,0.5113760192429467,0.5627279722016398,0.521384383082923,0.4915885617349205,0.5547502683415307
+full_training_164,Exp. 164,2000,1000,1000,flipped,0.5366759999999999,0.5120504276136731,0.561220933977117,0.5161446841568343,0.4858728031177736,0.5473552680228567
+full_training_163,Exp. 163,2000,1000,1000,flipped,0.530806,0.5065650618387278,0.5559242743945728,0.5286159101873922,0.5002866956247786,0.5615049195410815
+full_training_158,Exp. 158,2000,1000,1000,flipped,0.522809,0.49770570828257793,0.5475559067173488,0.505955734623714,0.47701996626172605,0.537548312209683
+full_training_57,Exp. 57,2000,1000,1000,flipped,0.52255,0.4972441361371893,0.5478754914582684,0.5174217405439884,0.48837962237884286,0.5495949460185908
+full_training_156,Exp. 156,2000,1000,1000,flipped,0.5212655,0.49722371198630455,0.5466177083196124,0.514623817289043,0.4860649744300327,0.5459847906712552
+full_training_160,Exp. 160,2000,1000,1000,flipped,0.518996,0.49417088468034387,0.5443120056153771,0.505881547655754,0.47767294163006,0.5372619710489989
+full_training_03,Exp. 3,2000,1000,1000,raw,0.516398,0.49087888813360325,0.5411363252785548,0.4987789336774834,0.47006638796966244,0.5297683529654459
+full_training_151,Exp. 151,2000,1000,1000,flipped,0.5069889999999999,0.4818475395874359,0.5316817642760632,0.4933123868505951,0.46432125076824693,0.5235070856422964
+full_training_153,Exp. 153,2000,1000,1000,flipped,0.5069889999999999,0.4818475395874359,0.5316817642760632,0.4933123868505951,0.46432125076824693,0.5235070856422964
+full_training_152,Exp. 152,2000,1000,1000,flipped,0.5069889999999999,0.4818475395874359,0.5316817642760632,0.4933123868505951,0.46432125076824693,0.5235070856422964
+full_training_145,Exp. 145,2000,1000,1000,flipped,0.502603,0.4777915746628979,0.5282889030942474,0.4989344793257081,0.47139907418293064,0.5306550878783755
+full_training_149,Exp. 149,2000,1000,1000,flipped,0.502603,0.4777915746628979,0.5282889030942474,0.4989344793257081,0.47139907418293064,0.5306550878783755

thesis_repro/phylo_msa1_outputs/VEP_ALL_BASE_MODELS_STRICT_METRICS.csv

@@ -0,0 +1,176 @@
+ablation_id,in_ablation_registry,class,configuration,checkpoint_stem,n,pos,neg,auroc,auprc,auroc_flipped,auprc_flipped,auroc_calibrated,auprc_calibrated,polarity,score_mean,score_std,branch_tag
+1,True,Baseline,"Baseline (Full Configuration: MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_01,247,47,200,0.5925531914893617,0.29646943161175754,0.40744680851063825,0.17613683146966158,0.5925531914893617,0.29646943161175754,raw,-0.027529910025808974,0.4209074331846881,legacy 7-way line
+2,True,Component Removal,"Component Removal: w/o Flip Nonconserved (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_02,247,47,200,0.54574468085106387,0.27306676471207214,0.45425531914893613,0.20147034330329708,0.54574468085106387,0.27306676471207214,raw,-0.14923071523426998,0.48065530671894047,legacy 7-way line
+3,True,Component Removal,"Component Removal: w/o Auxiliary Features (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_03,247,47,200,0.54946808510638301,0.24689463759122063,0.45053191489361699,0.18720254978461337,0.54946808510638301,0.24689463759122063,raw,-0.11860275630526214,0.34775809656629536,legacy 7-way line
+4,True,Component Removal,"Component Removal: w/o Soft Masked Loss (MLM 0.15, LR 0.0002, 512/8/6)",full_training_04,247,47,200,0.5669148936170213,0.27471619410519665,0.4330851063829787,0.19293349524346726,0.5669148936170213,0.27471619410519665,raw,-0.16007641162949535,0.51470816233217898,legacy 7-way line
+5,True,Component Removal,"Component Removal: w/o Conservation Weight (MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_05,247,47,200,0.56425531914893612,0.27529808081259849,0.43574468085106388,0.1710682155395887,0.56425531914893612,0.27529808081259849,raw,-0.14908715346564166,0.44437527540129529,legacy 7-way line
+6,True,Hyperparameter: MLM,"Hyperparameter: MLM 0.1 (Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_06,247,47,200,0.56585106382978723,0.27516102300718942,0.43414893617021283,0.17521298597753693,0.56585106382978723,0.27516102300718942,raw,-0.074615198832291826,0.45149482034430866,legacy 7-way line
+7,True,Hyperparameter: MLM,"Hyperparameter: MLM 0.2 (Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_07,247,47,200,0.52851063829787237,0.23972435195727221,0.47148936170212769,0.18738186547100139,0.52851063829787237,0.23972435195727221,raw,-0.15849626908900766,0.47220255696124602,legacy 7-way line
+8,True,Hyperparameter: Soft Mask,"Hyperparameter: Soft Mask 0.05 (MLM 0.15, LR 0.0002, 512/8/6)",full_training_08,247,47,200,0.60223404255319146,0.33136605729346347,0.39776595744680848,0.16630726327381373,0.60223404255319146,0.33136605729346347,raw,-0.099077157525398474,0.48523725161245634,legacy 7-way line
+9,True,Hyperparameter: Soft Mask,"Hyperparameter: Soft Mask 0.2 (MLM 0.15, LR 0.0002, 512/8/6)",full_training_09,247,47,200,0.58680851063829786,0.29707957313058941,0.41319148936170214,0.1740212843836213,0.58680851063829786,0.29707957313058941,raw,-0.022779455370748575,0.50579712055185944,legacy 7-way line
+10,True,Hyperparameter: Learning Rate,"Hyperparameter: Learning Rate 0.0001 (MLM 0.15, Soft Mask 0.1, 512/8/6)",full_training_10,247,47,200,0.57840425531914885,0.26858833747000466,0.42159574468085104,0.18920851868052352,0.57840425531914885,0.26858833747000466,raw,-0.079431968299966113,0.42325703059994529,legacy 7-way line
+100,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.085: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_100_64bp_MLM_0.085,247,47,200,0.46468085106382984,0.19229835410682622,0.53531914893617016,0.21911827750181592,0.53531914893617016,0.21911827750181592,flipped,-0.26338293870933627,0.59812030339411981,legacy 7-way line
+101,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + LR=0.00018: MLM 0.08, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_101_64bp_LR_0.00018,247,47,200,0.46829787234042558,0.27400955347237382,0.53170212765957447,0.23554913078809198,0.53170212765957447,0.27400955347237382,flipped,-0.1695632089970083,0.40167514592600484,legacy 7-way line
+102,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + LR=0.00022: MLM 0.08, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_102_64bp_LR_0.00022,247,47,200,0.47053191489361701,0.28416102920911307,0.52946808510638299,0.22621868670278289,0.52946808510638299,0.28416102920911307,flipped,-0.2036176982678865,0.46254644413349044,legacy 7-way line
+103,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=750: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_103_64bp_warmup_750,247,47,200,0.48755319148936166,0.27543544572087225,0.51244680851063829,0.20928996760567367,0.51244680851063829,0.27543544572087225,flipped,-0.22739928454040034,0.45477677622675483,legacy 7-way line
+104,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=1250: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_104_64bp_warmup_1250,247,47,200,0.40776595744680849,0.1916535273929913,0.59223404255319145,0.26862008133675991,0.59223404255319145,0.26862008133675991,flipped,-0.24364200391267476,0.54493405235572856,legacy 7-way line
+105,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + warmup=1500: MLM 0.08, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_105_64bp_warmup_1500,247,47,200,0.47063829787234041,0.26612369960462612,0.52936170212765965,0.2235122705826911,0.52936170212765965,0.26612369960462612,flipped,-0.26405701897887568,0.46932974320816823,legacy 7-way line
+106,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + soft_masked_loss_weight=0.15: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_106_64bp_soft_mask_0.15,247,47,200,0.50957446808510631,0.26170915147860563,0.49042553191489358,0.20032732985047241,0.50957446808510631,0.26170915147860563,raw,-0.17288849469621173,0.44221185252265727,legacy 7-way line
+107,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 类别权重Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_107_64bp_class_weights,247,47,200,0.38085106382978728,0.16243110659116514,0.61914893617021272,0.31024621901749583,0.61914893617021272,0.31024621901749583,flipped,-0.33428882913068236,0.47840901089623278,legacy 7-way line
+108,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True (alpha=0.25, gamma=2.0), eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_108_64bp_focal_loss,247,47,200,0.48617021276595745,0.27471267524494791,0.5138297872340426,0.21064439281126168,0.5138297872340426,0.27471267524494791,flipped,-0.15299852126040439,0.33110900119452952,legacy 7-way line
+109,True,Model Capacity: 64bp,"Exp 80优化 + 64bp窗口 + 768/12/8架构: MLM 0.08, LR 0.0002, cosine, warmup 1000, 768/12/8, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_109_64bp_768_12_8,247,47,200,0.50787234042553187,0.25259710355760495,0.49212765957446808,0.19146522792252926,0.50787234042553187,0.25259710355760495,raw,-0.13020200936900458,0.360101125359236,legacy 7-way line
+11,True,Hyperparameter: Learning Rate,"Hyperparameter: Learning Rate 0.0003 (MLM 0.15, Soft Mask 0.1, 512/8/6)",full_training_11,247,47,200,0.56659574468085105,0.28256559145660443,0.43340425531914895,0.19154597742272561,0.56659574468085105,0.28256559145660443,raw,-0.1597396804977525,0.54797594867597832,legacy 7-way line
+110,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 更激进类别权重Loss: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True (aggressive), eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_110_64bp_class_weights_aggressive,247,47,200,0.35670212765957449,0.14628401278201408,0.64329787234042546,0.28318614649979973,0.64329787234042546,0.28318614649979973,flipped,-0.53930251945850816,0.6440031973777236,legacy 7-way line
+111,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss (alpha=0.5, gamma=1.5): MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_111_64bp_focal_loss_tuned,247,47,200,0.49191489361702134,0.28032011842424809,0.50808510638297855,0.22510695778647638,0.50808510638297855,0.28032011842424809,flipped,-0.13070857283557474,0.3492132427871264,legacy 7-way line
+112,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + 组合优化（类别权重 + Focal Loss）: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_class_weights=True, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_112_64bp_combined_loss,247,47,200,0.42936170212765962,0.22692939767349091,0.57063829787234044,0.23633616364034399,0.57063829787234044,0.23633616364034399,flipped,-0.36579222862537092,0.48547353059630355,legacy 7-way line
+113,True,Loss Optimization: 64bp,"Exp 80优化 + 64bp窗口 + Focal Loss (alpha=0.25, gamma=2.5): MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, use_focal_loss=True, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_113_64bp_focal_loss_gamma25,247,47,200,0.50170212765957445,0.26908343445059402,0.49829787234042555,0.21083985061483534,0.50170212765957445,0.26908343445059402,raw,-0.10321414518935478,0.26557026787226751,legacy 7-way line
+114,True,GPN-MSA: Original Implementation,"GPN-MSA (Original Implementation): MLM 0.15, LR 0.0001, cosine scheduler, warmup 1000, 512/8/6, n_aux=30, batch_size=512×1",full_training_114_64bp_MLM_0.08_retrain,247,47,200,0.49787234042553191,0.22154087320831706,0.50212765957446803,0.24186517759529338,0.50212765957446803,0.24186517759529338,flipped,-0.22597313265086186,0.54483863098017571,legacy 7-way line
+115,True,Masking Ratio,"Masking Ratio: MLM 0.30 (基于Exp 10: 128bp + aux_features, LR 0.0001, batch=2048)",full_training_115,247,47,200,0.49170212765957444,0.20672348631484688,0.50829787234042545,0.20936247017451698,0.50829787234042545,0.20936247017451698,flipped,-0.24009886613258949,0.39761707206516922,legacy 7-way line
+116,True,Masking Ratio,"Masking Ratio: MLM 0.50 (基于Exp 10: 128bp + aux_features, LR 0.0001, batch=2048)",full_training_116,247,47,200,0.48106382978723411,0.20932974883564714,0.51893617021276595,0.20092239638040466,0.51893617021276595,0.20932974883564714,flipped,-0.3045068829648408,0.44384518193475841,legacy 7-way line
+117,True,Block/Span Masking,"Span Masking: mean_span_length=3.0 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_117,247,47,200,0.4898936170212766,0.20487591058881843,0.51010638297872335,0.20747607465161272,0.51010638297872335,0.20747607465161272,flipped,-0.2147802086132258,0.37925001536988306,legacy 7-way line
+118,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_118,247,47,200,0.47734042553191491,0.19631622107151875,0.52265957446808509,0.20820645286708081,0.52265957446808509,0.20820645286708081,flipped,-0.25273241132859758,0.39175218119026894,legacy 7-way line
+119,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+A1 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_119,247,47,200,0.51946808510638298,0.22733535033930716,0.48053191489361696,0.18796823456764139,0.51946808510638298,0.22733535033930716,raw,-0.20404887362289043,0.41775923882062538,legacy 7-way line
+12,True,Architecture,"Architecture: 384/8/6 (Small Model, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_12,247,47,200,0.55297872340425536,0.26489968329820213,0.44702127659574464,0.19341098845345678,0.55297872340425536,0.26489968329820213,raw,-0.098899822486074346,0.42979134798442481,legacy 7-way line
+120,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+A1+A2 (基于Exp 10: 128bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_120,247,47,200,0.48234042553191492,0.20009723394145096,0.51765957446808508,0.20546651276027666,0.51765957446808508,0.20546651276027666,flipped,-0.26412185568075913,0.44244224932161902,legacy 7-way line
+121,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 64bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_121,247,47,200,0.48638297872340425,0.25410449312684125,0.51361702127659581,0.27067711344061834,0.51361702127659581,0.27067711344061834,flipped,-0.17299634169953071,0.50880827019786401,legacy 7-way line
+122,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 256bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_122,247,47,200,0.37,0.147387787376632,0.63,0.30421203373718181,0.63,0.30421203373718181,flipped,-0.45672186002557574,0.53149423015800645,legacy 7-way line
+123,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_123,247,47,200,0.5562765957446808,0.27589234508597371,0.44372340425531909,0.18742690799582482,0.5562765957446808,0.27589234508597371,raw,-0.15702775603363872,0.39475887788776953,legacy 7-way line
+124,True,Block/Span Masking,"Span Masking: mean_span_length=6.0 (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_124,247,47,200,0.44351063829787241,0.20496496435142381,0.55648936170212771,0.24157516439581009,0.55648936170212771,0.24157516439581009,flipped,-0.17079269644702494,0.25315131354447401,legacy 7-way line
+125,True,Block/Span Masking,"Span Masking: mean_span_length=12.0 (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_125,247,47,200,0.44712765957446809,0.22489645249141949,0.55287234042553191,0.22326713778005455,0.55287234042553191,0.22489645249141949,flipped,-0.10299530685671911,0.189017206219668,legacy 7-way line
+126,True,Window Size: 256bp,"Window Size: 256bp (MLM 0.15, LR 0.0001, batch=2048, 512/8/12)",full_training_126,247,47,200,0.61989361702127654,0.34113249632616094,0.3801063829787234,0.18241353140132255,0.61989361702127654,0.34113249632616094,raw,0.16651838076742073,0.5912770837839757,legacy window sweep
+127,True,Window Size: 512bp,"Window Size: 512bp (MLM 0.15, LR 0.0001, batch=2048, 768/12/6)",full_training_127,247,47,200,0.3681914893617021,0.14805272237455752,0.6318085106382979,0.25866070421122472,0.6318085106382979,0.25866070421122472,flipped,-0.49049361852499157,0.70684083652885865,legacy window sweep
+128,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0001, batch=128, 768/12/12)",full_training_128,247,47,200,0.49117021276595735,0.2241696673369819,0.5088297872340426,0.23790865853964693,0.5088297872340426,0.23790865853964693,flipped,-0.20181757525393837,0.36758573152244145,legacy window sweep
+129,True,Window Size: 64bp,"Window Size: 64bp (MLM 0.15, LR 0.0001, batch=2048, 768/12/12)",full_training_129,247,47,200,0.48372340425531912,0.28408803344425637,0.51627659574468088,0.22391021514535719,0.51627659574468088,0.28408803344425637,flipped,-0.14573394142181767,0.47260610639909129,legacy window sweep
+13,True,Architecture,"Architecture: 768/12/8 (Large Model, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_13,247,47,200,0.55840425531914895,0.2666251081952124,0.44159574468085105,0.17818397179950923,0.55840425531914895,0.2666251081952124,raw,-0.14330200754437852,0.4726721964594548,legacy 7-way line
+130,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 256bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_130,247,47,200,0.38744680851063829,0.16699247802726258,0.61255319148936171,0.23284266288950312,0.61255319148936171,0.23284266288950312,flipped,-0.67535875749732799,0.82431048344330216,legacy 7-way line
+131,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_131,247,47,200,0.5288297872340425,0.24644192144386298,0.47117021276595744,0.19341886321686583,0.5288297872340425,0.24644192144386298,raw,-0.28913306707312703,0.44169089187231564,legacy 7-way line
+132,True,Synchronized Ancestral Masking,"Synchronized Masking: Human+Chimp (基于Exp 10: 1024bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_132,247,47,200,0.43659574468085105,0.19362498636378417,0.56340425531914884,0.24001937907003434,0.56340425531914884,0.24001937907003434,flipped,-0.22111769641941859,0.2979178718506143,legacy 7-way line
+133,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0003, batch=2048, 512/8/6)",full_training_133,247,47,200,0.45446808510638304,0.20339990297684041,0.54553191489361708,0.24125664184584145,0.54553191489361708,0.24125664184584145,flipped,-0.20692495875030395,0.31044521940208669,legacy window sweep
+134,True,Window Size: 1024bp,"Window Size: 1024bp (MLM 0.15, LR 0.0001, batch=2048, 512/8/6, warmup=2000)",full_training_134,247,47,200,0.4503191489361702,0.20808949710247648,0.5496808510638298,0.23733802239968696,0.5496808510638298,0.23733802239968696,flipped,-0.15322031521121501,0.2350950071968583,legacy window sweep
+135,True,Loss Optimization,"Focal Loss: gamma=1.0 (基于Exp 10: 512bp + aux_features, MLM 0.15, LR 0.0001, batch=2048)",full_training_135,247,47,200,0.54638297872340424,0.27782422670318374,0.4536170212765957,0.19389437961126227,0.54638297872340424,0.27782422670318374,raw,-0.16711359492197694,0.38242257758553366,legacy 7-way line
+136,True,特定序列Masking,"Exp 136 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_136,247,47,200,0.47574468085106386,0.20490352046175561,0.52425531914893608,0.19885676451995593,0.52425531914893608,0.20490352046175561,flipped,-0.32318984496931313,0.48158712317634467,legacy 7-way line
+137,True,特定序列Masking,"Exp 137 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_137,247,47,200,0.56095744680851056,0.31971270877634594,0.43904255319148938,0.2052289758742275,0.56095744680851056,0.31971270877634594,raw,-0.13978164446981331,0.53784961624908578,legacy 7-way line
+138,True,特定序列Masking,"Exp 138 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_138,247,47,200,0.48425531914893616,0.20923834311724687,0.51574468085106373,0.20701429083347772,0.51574468085106373,0.20923834311724687,flipped,-0.16433197211640083,0.44223640389534907,legacy 7-way line
+139,True,特定序列Masking,"Exp 139 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_139,247,47,200,0.51776595744680853,0.23085423187711765,0.48223404255319152,0.18886667367579876,0.51776595744680853,0.23085423187711765,raw,-0.21155490795610404,0.45436925148342444,legacy 7-way line
+14,True,Optimized Combination,Optimized Combination: MLM 0.2 + Soft Mask 0.2 + LR 0.0001 (512/8/6),full_training_14,247,47,200,0.61148936170212764,0.313923492076564,0.38851063829787236,0.17715326378000168,0.61148936170212764,0.313923492076564,raw,-0.0073821347976020472,0.43208175133455379,legacy 7-way line
+140,True,特定序列Masking,"Exp 140 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_140,247,47,200,0.47446808510638294,0.20274687910389994,0.52553191489361706,0.20322086841607884,0.52553191489361706,0.20322086841607884,flipped,-0.33823515758340655,0.47781772890332291,legacy 7-way line
+141,True,特定序列Masking,"Exp 141 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_141,247,47,200,0.50595744680851062,0.22856044037196496,0.49404255319148926,0.19214095266807446,0.50595744680851062,0.22856044037196496,raw,-0.19181640910716191,0.48617453354071577,legacy 7-way line
+142,True,特定序列Masking,"Exp 142 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_142,247,47,200,0.44223404255319149,0.21770418326077745,0.55776595744680846,0.24862001935286915,0.55776595744680846,0.24862001935286915,flipped,-0.33053404285840177,0.54639492283690838,legacy 7-way line
+143,True,特定序列Masking,"Exp 143 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_143,247,47,200,0.52308510638297867,0.25982105517518239,0.47691489361702122,0.23163686236407499,0.52308510638297867,0.25982105517518239,raw,-0.18238665893492911,0.53254412220766889,legacy 7-way line
+144,True,特定序列Masking,"Exp 144 (特定序列Masking): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_144,247,47,200,0.53680851063829782,0.27353037924632073,0.46319148936170212,0.19460712001139652,0.53680851063829782,0.27353037924632073,raw,-0.17301026541694456,0.38348798172456833,legacy 7-way line
+145,True,Auxiliary Features数量探索,"Exp 145 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=3, 128bp",full_training_145,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+146,True,Auxiliary Features数量探索,"Exp 146 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=6, 128bp",full_training_146,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+147,True,Auxiliary Features数量探索,"Exp 147 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=9, 128bp",full_training_147,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+148,True,Auxiliary Features数量探索,"Exp 148 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=12, 128bp",full_training_148,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+149,True,Auxiliary Features数量探索,"Exp 149 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_149,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+15,True,Optimized Combination,Optimized Combination: MLM 0.2 + Soft Mask 0.2 + LR 0.0001 (768/12/8),full_training_15,247,47,200,0.61234042553191492,0.27488224587084992,0.38765957446808508,0.15248369008433174,0.61234042553191492,0.27488224587084992,raw,-0.068816064099068583,0.59842791633695303,legacy 7-way line
+150,True,Auxiliary Features数量探索,"Exp 150 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=18, 128bp",full_training_150,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+151,True,Auxiliary Features数量探索,"Exp 151 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=21, 128bp",full_training_151,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+152,True,Auxiliary Features数量探索,"Exp 152 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=24, 128bp",full_training_152,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+153,True,Auxiliary Features数量探索,"Exp 153 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=27, 128bp",full_training_153,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+154,True,Auxiliary Features数量探索,"Exp 154 (Auxiliary Features数量探索): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 128bp",full_training_154,247,47,200,0.49180851063829784,0.19840646591917618,0.50819148936170211,0.25317322934334796,0.50819148936170211,0.25317322934334796,flipped,-0.12009834277967692,0.44607233212764758,legacy 7-way line
+155,True,组合策略,"Exp 155 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_155,247,47,200,0.47574468085106386,0.20490352046175561,0.52425531914893608,0.19885676451995593,0.52425531914893608,0.20490352046175561,flipped,-0.32318984496931313,0.48158712317634467,legacy 7-way line
+156,True,组合策略,"Exp 156 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_156,247,47,200,0.56095744680851056,0.31971270877634594,0.43904255319148938,0.2052289758742275,0.56095744680851056,0.31971270877634594,raw,-0.13978164446981331,0.53784961624908578,legacy 7-way line
+157,True,组合策略,"Exp 157 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=9, 128bp",full_training_157,247,47,200,0.47574468085106386,0.20490352046175561,0.52425531914893608,0.19885676451995593,0.52425531914893608,0.20490352046175561,flipped,-0.32318984496931313,0.48158712317634467,legacy 7-way line
+158,True,组合策略,"Exp 158 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=21, 128bp",full_training_158,247,47,200,0.47574468085106386,0.20490352046175561,0.52425531914893608,0.19885676451995593,0.52425531914893608,0.20490352046175561,flipped,-0.32318984496931313,0.48158712317634467,legacy 7-way line
+159,True,组合策略,"Exp 159 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 128bp",full_training_159,247,47,200,0.50595744680851062,0.22856044037196496,0.49404255319148926,0.19214095266807446,0.50595744680851062,0.22856044037196496,raw,-0.19181640910716191,0.48617453354071577,legacy 7-way line
+16,True,Optimized Combination,Optimized Combination: MLM 0.1 + Soft Mask 0.2 + LR 0.0001 (512/8/6),full_training_16,247,47,200,0.55925531914893623,0.26259166940305945,0.44074468085106389,0.19744803234341723,0.55925531914893623,0.26259166940305945,raw,-0.093236754901013394,0.40668303526914845,legacy 7-way line
+160,True,组合策略,"Exp 160 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_160,247,47,200,0.52308510638297867,0.25982105517518239,0.47691489361702122,0.23163686236407499,0.52308510638297867,0.25982105517518239,raw,-0.18238665893492911,0.53254412220766889,legacy 7-way line
+161,True,组合策略,"Exp 161 (组合策略): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_161,247,47,200,0.53680851063829782,0.27353037924632073,0.46319148936170212,0.19460712001139652,0.53680851063829782,0.27353037924632073,raw,-0.17301026541694456,0.38348798172456833,legacy 7-way line
+162,True,基于最佳配置的优化,"Exp 162 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_162,247,47,200,0.45478723404255322,0.2692296743149476,0.54521276595744683,0.26715002048649017,0.54521276595744683,0.2692296743149476,flipped,-0.21775774820613475,0.51065807167957145,legacy 7-way line
+163,True,基于最佳配置的优化,"Exp 163 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_163,247,47,200,0.49670212765957444,0.26990957078651079,0.50329787234042545,0.22565640856503422,0.50329787234042545,0.26990957078651079,flipped,-0.10258202321133633,0.49081946874389581,legacy 7-way line
+164,True,基于最佳配置的优化,"Exp 164 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_164,247,47,200,0.45478723404255322,0.2692296743149476,0.54521276595744683,0.26715002048649017,0.54521276595744683,0.2692296743149476,flipped,-0.21775774820613475,0.51065807167957145,legacy 7-way line
+165,True,基于最佳配置的优化,"Exp 165 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_165,247,47,200,0.53680851063829782,0.27353037924632073,0.46319148936170212,0.19460712001139652,0.53680851063829782,0.27353037924632073,raw,-0.17301026541694456,0.38348798172456833,legacy 7-way line
+166,True,基于最佳配置的优化,"Exp 166 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 512bp, Focal Loss (α=0.25, γ=1.0)",full_training_166,247,47,200,0.54638297872340424,0.27782422670318374,0.4536170212765957,0.19389437961126227,0.54638297872340424,0.27782422670318374,raw,-0.16711359492197694,0.38242257758553366,legacy 7-way line
+167,True,基于最佳配置的优化,"Exp 167 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 64bp",full_training_167,247,47,200,0.52308510638297867,0.25982105517518239,0.47691489361702122,0.23163686236407499,0.52308510638297867,0.25982105517518239,raw,-0.18238665893492911,0.53254412220766889,legacy 7-way line
+168,True,基于最佳配置的优化,"Exp 168 (基于最佳配置的优化): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=15, 64bp",full_training_168,247,47,200,0.57925531914893624,0.31127554695304294,0.42074468085106381,0.21837948090207926,0.57925531914893624,0.31127554695304294,raw,-0.016213536745140911,0.59489292243021785,legacy 7-way line
+169,True,10-primates数据集,Exp 169 (see config full_training_169.json),full_training_169,247,47,200,0.31968085106382982,0.13556538631519807,0.68031914893617018,0.26902798245778337,0.68031914893617018,0.26902798245778337,flipped,-5.6751815259215324,2.6620200594401386,10-primate (early)
+17,True,Architecture,"Architecture: 768/8/6 (Hidden Size Only, MLM 0.15, Soft Mask 0.1, LR 0.0002)",full_training_17,247,47,200,0.55287234042553191,0.33082124384019607,0.44712765957446809,0.19420254247736737,0.55287234042553191,0.33082124384019607,raw,-0.062094009839571439,0.50002678227502773,legacy 7-way line
+18,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_18,247,47,200,0.58840425531914897,0.32356548117572909,0.41159574468085108,0.1990727194753211,0.58840425531914897,0.32356548117572909,raw,-0.0028065190865443302,0.52084662067542908,legacy 7-way line
+19,True,Optimal Combination,Optimal Combination: MLM 0.2 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_19,247,47,200,0.62223404255319159,0.32382823618781403,0.37776595744680846,0.18560544657574077,0.62223404255319159,0.32382823618781403,raw,0.066541971525682614,0.51783882038361972,legacy 7-way line
+20,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.1 + LR 0.0002,full_training_20,247,47,200,0.58712765957446811,0.32160805643385726,0.41287234042553189,0.18178405875020764,0.58712765957446811,0.32160805643385726,raw,0.012524824272765805,0.49157897287900448,legacy 7-way line
+201,True,10-primates新数据集: 1024bp基准,"Exp 201 (新10-primates数据集): MLM 0.15, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_201,247,47,200,0.32329787234042551,0.13754590440436817,0.67670212765957449,0.3315454086452943,0.67670212765957449,0.3315454086452943,flipped,-2.2137599696031951,2.0281767662906818,new 10-primate
+202,True,10-primates新数据集: MLM 0.1,"Exp 202 (新10-primates数据集): MLM 0.1, Soft Mask 0.1, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_202,247,47,200,0.3299999999999999,0.13846924778216596,0.67000000000000004,0.33264783014758115,0.67000000000000004,0.33264783014758115,flipped,-2.6569749302709633,2.3790492302501129,new 10-primate
+203,True,10-primates新数据集: Soft Mask 0.2,"Exp 203 (新10-primates数据集): MLM 0.15, Soft Mask 0.2, LR 0.0001, 512/8/6, n_aux=30, 1024bp",full_training_203,247,47,200,0.3234042553191489,0.13755963314292202,0.67659574468085115,0.33075201376341801,0.67659574468085115,0.33075201376341801,flipped,-2.2039898522952308,2.0192676674908276,new 10-primate
+204,True,10-primates新数据集: LR 0.0002,"Exp 204 (新10-primates数据集): MLM 0.15, Soft Mask 0.1, LR 0.0002, 512/8/6, n_aux=90, 1024bp",full_training_204,247,47,200,0.33117021276595748,0.13845862719741317,0.66882978723404252,0.33217599854074764,0.66882978723404252,0.33217599854074764,flipped,-2.5523211031790205,2.3120423348478103,new 10-primate
+206,True,10-primates新数据集,Exp 206 (see config full_training_206.json),full_training_206,247,47,200,0.3680851063829787,0.14454234642076971,0.63191489361702124,0.25610538252562493,0.63191489361702124,0.25610538252562493,flipped,-5.2701826742303517,2.5125715003443205,new 10-primate
+207,True,10-primates新数据集,Exp 207 (see config full_training_207.json),full_training_207,247,47,200,0.34170212765957442,0.13943723855017542,0.65829787234042558,0.30332713880961437,0.65829787234042558,0.30332713880961437,flipped,-4.4111963153368068,2.5653975236785365,new 10-primate
+209,False,Extended run,Experiment 209: checkpoint in outputs/full_training_209; not yet in ABLATION_SUMMARY.csv,full_training_209,247,47,200,0.34500000000000003,0.14042322317653463,0.65500000000000003,0.28096100918964045,0.65500000000000003,0.28096100918964045,flipped,-4.6116000230495748,2.4562047671529688,new 10-primate
+21,True,Optimal Combination,Optimal Combination: MLM 0.1 + 768/8/6 + Soft Mask 0.2 + LR 0.0002,full_training_21,247,47,200,0.5585106382978724,0.31844606643540085,0.44148936170212771,0.19892828869608631,0.5585106382978724,0.31844606643540085,raw,-0.014880998776509212,0.42933454686693717,legacy 7-way line
+210,False,Extended run,Experiment 210: checkpoint in outputs/full_training_210; not yet in ABLATION_SUMMARY.csv,full_training_210,247,47,200,0.32638297872340427,0.13730717563419614,0.67361702127659573,0.30297281269896004,0.67361702127659573,0.30297281269896004,flipped,-2.7600071685999512,2.5179756694675932,new 10-primate
+22,True,Optimal Combination,Optimal Combination: MLM 0.15 + 768/8/6 + Soft Mask 0.2 + LR 0.0001,full_training_22,247,47,200,0.64606382978723409,0.35218228109334326,0.35393617021276591,0.15997522008703288,0.64606382978723409,0.35218228109334326,raw,0.11551677396422938,0.61015476441818361,legacy 7-way line
+23,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 (768/8/6),full_training_23,247,47,200,0.49138297872340425,0.26812264620079596,0.50861702127659569,0.24434613523602866,0.50861702127659569,0.26812264620079596,flipped,-0.093898243629015415,0.42796332240598783,legacy 7-way line
+24,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.2 + LR 0.0002 (512/8/6),full_training_24,247,47,200,0.60265957446808516,0.28587580294524362,0.3973404255319149,0.19090602555591812,0.60265957446808516,0.28587580294524362,raw,-0.055597194172592783,0.50532754336592789,legacy 7-way line
+25,True,Optimal Combination,Optimal Combination: MLM 0.1 + Soft Mask 0.1 + LR 0.00015 (512/8/6),full_training_25,247,47,200,0.56861702127659564,0.30649549905750706,0.43138297872340425,0.20172833218982805,0.56861702127659564,0.30649549905750706,raw,-0.070320025086402893,0.43966106327440557,legacy 7-way line
+26,True,Improvement,Improvement: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 (768/8/6),full_training_26,247,47,200,0.60999999999999999,0.32194126644398047,0.38999999999999996,0.1665662855884778,0.60999999999999999,0.32194126644398047,raw,-0.025531555205462915,0.50271773039058598,legacy 7-way line
+27,True,Improvement,Improvement: MLM 0.1 + Soft Mask 0.1 + LR 0.0002 + Cosine LR (512/8/6),full_training_27,247,47,200,0.54510638297872338,0.26383669796930542,0.45489361702127656,0.18246188149572046,0.54510638297872338,0.26383669796930542,raw,-0.11910052265715503,0.42751838034997686,legacy 7-way line
+28,True,Architecture: Heads,"Architecture: 480/12/6 (Heads Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_28,247,47,200,0.47829787234042548,0.23767021480570952,0.52170212765957436,0.2191955091134189,0.52170212765957436,0.23767021480570952,flipped,-0.20913464072262228,0.42813754085054545,legacy 7-way line
+29,True,Architecture: Heads,"Architecture: 512/16/6 (Heads Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_29,247,47,200,0.45787234042553193,0.20889290480114769,0.54212765957446807,0.2125763960254401,0.54212765957446807,0.2125763960254401,flipped,-0.28403334476445852,0.44774742334390844,legacy 7-way line
+30,True,Architecture: Layers,"Architecture: 512/8/8 (Layers Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_30,247,47,200,0.5875531914893618,0.33582064722142357,0.41244680851063831,0.19036846664699342,0.5875531914893618,0.33582064722142357,raw,-0.069152546918343924,0.41857885657425525,legacy 7-way line
+31,True,Architecture: Layers,"Architecture: 512/8/12 (Layers Only, MLM 0.1, Soft Mask 0.1, LR 0.0002)",full_training_31,247,47,200,0.51265957446808508,0.29119714812383418,0.48734042553191498,0.187276804939145,0.51265957446808508,0.29119714812383418,raw,-0.19530953484990818,0.43846785331429527,legacy 7-way line
+32,True,Learning Rate,"Learning Rate: LR 0.00015 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_32,247,47,200,0.56659574468085105,0.26139115765869481,0.43340425531914895,0.20618304749301752,0.56659574468085105,0.26139115765869481,raw,-0.046250312585338405,0.39155527815137797,legacy 7-way line
+33,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_33,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+34,True,Learning Rate,"Learning Rate: LR 0.00025 (cosine, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_34,247,47,200,0.50723404255319149,0.25859496424298062,0.49276595744680857,0.19104841558920893,0.50723404255319149,0.25859496424298062,raw,-0.25102929789045081,0.42385539673382877,legacy 7-way line
+35,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.05, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_35,247,47,200,0.50638297872340443,0.26035655803978652,0.49361702127659574,0.18949793200671949,0.50638297872340443,0.26035655803978652,raw,-0.23948510588421995,0.42599215045702415,legacy 7-way line
+36,True,Learning Rate,"Learning Rate: LR 0.0002 (cosine, warmup_ratio=0.15, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_36,247,47,200,0.51265957446808519,0.25604622700603946,0.48734042553191492,0.18731817595035527,0.51265957446808519,0.25604622700603946,raw,-0.25199547566865621,0.42195231139812117,legacy 7-way line
+37,True,Learning Rate,"Learning Rate: LR 0.0002 (polynomial, warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, 512/8/6)",full_training_37,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+38,True,Window Size,"Window Size: 64bp (MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_38,247,47,200,0.50159574468085111,0.2445740350615988,0.49840425531914895,0.20186427982314256,0.50159574468085111,0.2445740350615988,raw,-0.1671941965697748,0.39928718381720069,legacy window sweep
+39,True,Window Size,"Window Size: 256bp (MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6)",full_training_39,247,47,200,0.52170212765957436,0.30585386373046852,0.47829787234042559,0.22182808168382928,0.52170212765957436,0.30585386373046852,raw,-0.12147468557724586,0.36354552575770227,legacy window sweep
+40,True,Learning Rate: Cosine,"Exp 06配置 + cosine LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_40,247,47,200,0.58468085106382983,0.33362364368947972,0.41531914893617022,0.18937498084303467,0.58468085106382983,0.33362364368947972,raw,-0.080971530755521798,0.40908814454519987,legacy 7-way line
+41,True,Learning Rate: Cosine,"Exp 19配置 + cosine LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_41,247,47,200,0.59489361702127652,0.34188601294022014,0.40510638297872337,0.20638331400987814,0.59489361702127652,0.34188601294022014,raw,0.023890289217836945,0.47918615815077703,legacy 7-way line
+42,True,Learning Rate,"Exp 06配置 + polynomial LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_42,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+43,True,Learning Rate,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_43,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+44,True,Learning Rate,"Exp 06配置 + constant_with_warmup LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_44,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+45,True,Learning Rate,"Exp 06配置 + cosine LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_45,247,47,200,0.57978723404255317,0.33884881828615515,0.42021276595744678,0.20893370733713307,0.57978723404255317,0.33884881828615515,raw,0.0051585419578590854,0.38769357133734522,legacy 7-way line
+46,True,Learning Rate,"Exp 06配置 + polynomial LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_46,247,47,200,0.57978723404255317,0.33884881828615515,0.42021276595744678,0.20893370733713307,0.57978723404255317,0.33884881828615515,raw,0.0051585419578590854,0.38769357133734522,legacy 7-way line
+47,True,Learning Rate,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_47,247,47,200,0.57978723404255317,0.33884881828615515,0.42021276595744678,0.20893370733713307,0.57978723404255317,0.33884881828615515,raw,0.0051585419578590854,0.38769357133734522,legacy 7-way line
+48,True,Learning Rate,"Exp 19配置 + polynomial LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_48,247,47,200,0.60553191489361691,0.35311736369879498,0.39446808510638298,0.17722844502387813,0.60553191489361691,0.35311736369879498,raw,0.016722077179534233,0.48365147388901475,legacy 7-way line
+49,True,Learning Rate,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_49,247,47,200,0.60553191489361691,0.35311736369879498,0.39446808510638298,0.17722844502387813,0.60553191489361691,0.35311736369879498,raw,0.016722077179534233,0.48365147388901475,legacy 7-way line
+50,True,Learning Rate,"Exp 19配置 + constant_with_warmup LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_50,247,47,200,0.57925531914893624,0.28623779701908864,0.42074468085106381,0.17707871365284733,0.57925531914893624,0.28623779701908864,raw,0.0039669607694332414,0.43526323751310686,legacy 7-way line
+51,True,Learning Rate,"Exp 19配置 + cosine LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_51,247,47,200,0.57436170212765969,0.33491606971664473,0.42563829787234037,0.20386487982760707,0.57436170212765969,0.33491606971664473,raw,0.0055839590698118632,0.42821536412172534,legacy 7-way line
+52,True,Learning Rate,"Exp 19配置 + polynomial LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_52,247,47,200,0.57436170212765969,0.33491606971664473,0.42563829787234037,0.20386487982760707,0.57436170212765969,0.33491606971664473,raw,0.0055839590698118632,0.42821536412172534,legacy 7-way line
+53,True,Learning Rate,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.2, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_53,247,47,200,0.57436170212765969,0.33491606971664473,0.42563829787234037,0.20386487982760707,0.57436170212765969,0.33491606971664473,raw,0.0055839590698118632,0.42821536412172534,legacy 7-way line
+54,True,Ensemble,"Exp 06配置 + seed=42 (Ensemble member 1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_54,247,47,200,0.48968085106382975,0.2919221594697945,0.51031914893617025,0.19710252374454038,0.51031914893617025,0.2919221594697945,flipped,-0.26631205405300928,0.4411369340510542,legacy 7-way line
+55,True,Ensemble,"Exp 06配置 + seed=123 (Ensemble member 2, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_55,247,47,200,0.49340425531914889,0.28637009988676998,0.50659574468085111,0.20434453950095816,0.50659574468085111,0.28637009988676998,flipped,-0.14417072454927421,0.48536157712927241,legacy 7-way line
+56,True,Ensemble,"Exp 06配置 + seed=456 (Ensemble member 3, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_56,247,47,200,0.53180851063829793,0.26623417678090927,0.46819148936170213,0.18994631353343522,0.53180851063829793,0.26623417678090927,raw,-0.10346164213501008,0.48649633885593196,legacy 7-way line
+57,True,Priority A,"Architecture sweep: 768/8/6 (Exp6 hyperparams, MLM 0.1, Soft Mask 0.1, LR 0.0002, 128bp)",full_training_57,247,47,200,0.60734042553191492,0.29188352521884681,0.39265957446808514,0.20002312800789415,0.60734042553191492,0.29188352521884681,raw,0.0065909498616268762,0.32375990976632174,legacy 7-way line
+58,True,Priority A,"Architecture sweep: 512/8/8 (Exp6 hyperparams, MLM 0.1, Soft Mask 0.1, LR 0.0002, 128bp)",full_training_58,247,47,200,0.5875531914893618,0.33582064722142357,0.41244680851063831,0.19036846664699342,0.5875531914893618,0.33582064722142357,raw,-0.069152546918343924,0.41857885657425525,legacy 7-way line
+59,True,Priority A,"Regularization: weight_decay=0.05 (Exp6配置, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_59,247,47,200,0.51585106382978729,0.26277822558100816,0.48414893617021276,0.18980236726482641,0.51585106382978729,0.26277822558100816,raw,-0.21625421549144544,0.4106424252766086,legacy 7-way line
+60,True,Priority A,"Regularization: weight_decay=0.1 (Exp6配置, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_60,247,47,200,0.54808510638297869,0.26767515025587624,0.45191489361702131,0.18904813414782734,0.54808510638297869,0.26767515025587624,raw,-0.091395629622675625,0.3769622466106699,legacy 7-way line
+61,True,Priority A: LR Scheduler,"Exp 06配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.1, Soft Mask 0.1, LR 0.0002, 512/8/6, 128bp)",full_training_61,247,47,200,0.50808510638297877,0.25387142408469043,0.49191489361702123,0.19007787103302026,0.50808510638297877,0.25387142408469043,raw,-0.23068543218890666,0.38937836051217828,legacy 7-way line
+62,True,Priority A: LR Scheduler,"Exp 19配置 + linear LR scheduler (warmup_ratio=0.1, MLM 0.2, Soft Mask 0.2, LR 0.0001, 768/8/6, 128bp)",full_training_62,247,47,200,0.60553191489361691,0.35311736369879498,0.39446808510638298,0.17722844502387813,0.60553191489361691,0.35311736369879498,raw,0.016722077179534233,0.48365147388901475,legacy 7-way line
+65,True,Aux Features: All Species,"Exp 06 config + all 6 species as aux_features (MLM 0.1, LR 0.0002, 512/8/6, n_aux=30)",full_training_65,247,47,200,0.4317021276595745,0.17702381577818116,0.5682978723404255,0.22095462086650663,0.5682978723404255,0.22095462086650663,flipped,-0.41748449544192323,0.52895772324180934,legacy 7-way line
+68,True,Paper Reproduction + Eval Frequency Optimization,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=500)",full_training_68,247,47,200,0.47468085106382979,0.20036539527086683,0.52531914893617015,0.20677183893257886,0.52531914893617015,0.20677183893257886,flipped,-0.26170608799467204,0.45442447183616302,legacy 7-way line
+69,True,Paper Reproduction + Eval Frequency Optimization,"Paper hyperparams + n_aux=3 + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=3, eval_steps=500)",full_training_69,247,47,200,0.47468085106382979,0.20036539527086683,0.52531914893617015,0.20677183893257886,0.52531914893617015,0.20677183893257886,flipped,-0.26170608799467204,0.45442447183616302,legacy 7-way line
+70_seed123,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=123",full_training_70_paper_correct_seed123,247,47,200,0.49648936170212765,0.25371878811057336,0.50351063829787235,0.20349709469800958,0.50351063829787235,0.25371878811057336,flipped,-0.19550259972390857,0.46504652832157672,other
+70_seed42,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=42",full_training_70_paper_correct_seed42,247,47,200,0.49531914893617024,0.22477497929582063,0.50468085106382987,0.19996097274030908,0.50468085106382987,0.22477497929582063,flipped,-0.19791063316437879,0.37211198570070764,other
+70_seed456,True,Paper Reproduction: Multi-Seed,"Paper hyperparams + 所有6个species aux_features + 优化评估频率 (MLM 0.15, Soft Mask 0.1, LR 0.0001, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36, batch=2048), seed=456",full_training_70_paper_correct_seed456,247,47,200,0.43574468085106383,0.17793540603700553,0.56425531914893612,0.20702675388102726,0.56425531914893612,0.20702675388102726,flipped,-0.43980239060243614,0.60960573668684415,other
+71_seed123,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=123",full_training_71_optimized_seed123,247,47,200,0.50063829787234038,0.23437356296918174,0.49936170212765957,0.20180735976385553,0.50063829787234038,0.23437356296918174,raw,-0.26163370744419484,0.47475797049851731,other
+71_seed42,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=42",full_training_71_optimized_seed42,247,47,200,0.48925531914893616,0.2252906707253893,0.51074468085106384,0.20395674512502662,0.51074468085106384,0.2252906707253893,flipped,-0.2338616150593468,0.40446506338155719,other
+71_seed456,True,Exp 65 Optimized: Multi-Seed,"Exp 65优化: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=36 (1 epoch), batch=2048, seed=456",full_training_71_optimized_seed456,247,47,200,0.46265957446808514,0.2180725497670444,0.53734042553191486,0.21614201429241017,0.53734042553191486,0.2180725497670444,flipped,-0.2663862763146157,0.44078829020447818,other
+72,True,Window Size: 512bp,"Exp 71优化 + 512bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=9 (1 epoch), batch=2048, max_position_embeddings=512",full_training_72_512bp,247,47,200,0.49159574468085104,0.24080599913607972,0.50840425531914901,0.19961232361857592,0.50840425531914901,0.24080599913607972,flipped,-0.2829566255272159,0.38954968026391473,legacy window sweep
+73,True,Window Size: 1024bp,"Exp 71优化 + 1024bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=4 (1 epoch), batch=2048, max_position_embeddings=1024",full_training_73_1024bp,247,47,200,0.43936170212765951,0.19855343163603154,0.56063829787234043,0.24402950877490498,0.56063829787234043,0.24402950877490498,flipped,-0.17613868795425786,0.23155799236063923,legacy window sweep
+74,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_74_64bp,247,47,200,0.48670212765957444,0.26801491752931372,0.51329787234042556,0.25577787483988179,0.51329787234042556,0.26801491752931372,flipped,-0.19999836017245706,0.45400080684529309,legacy window sweep
+74_seed123,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64, seed=123",full_training_74_64bp_seed123,247,47,200,0.46872340425531916,0.19393767354103644,0.53127659574468078,0.2065476882209987,0.53127659574468078,0.2065476882209987,flipped,-0.25339595340041499,0.37424930254043459,legacy window sweep
+74_seed456,True,Window Size: 64bp,"Exp 71优化 + 64bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64, seed=456",full_training_74_64bp_seed456,247,47,200,0.45074468085106384,0.20277923341617268,0.54925531914893622,0.25513584199702588,0.54925531914893622,0.25513584199702588,flipped,-0.45857144753459977,0.79157449289442794,legacy window sweep
+75,True,Window Size: 256bp,"Exp 71优化 + 256bp窗口: MLM 0.1, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=18 (1 epoch), batch=2048, max_position_embeddings=256",full_training_75_256bp,247,47,200,0.39127659574468082,0.16832388984427393,0.60872340425531923,0.23493896606514036,0.60872340425531923,0.23493896606514036,flipped,-0.52177662912167999,0.68236363910129394,legacy window sweep
+76,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + LR=0.00015: MLM 0.1, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_76_64bp_LR_0.00015,247,47,200,0.48882978723404252,0.26218533152536072,0.51117021276595742,0.21551112313493587,0.51117021276595742,0.26218533152536072,flipped,-0.20173983583566149,0.4411661339243258,legacy 7-way line
+77,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + LR=0.00025: MLM 0.1, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_77_64bp_LR_0.00025,247,47,200,0.48340425531914893,0.23621966889910051,0.51659574468085112,0.22133690088537267,0.51659574468085112,0.23621966889910051,flipped,-0.35479910267509429,0.55024861161900818,legacy 7-way line
+78,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + warmup=500: MLM 0.1, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_78_64bp_warmup_500,247,47,200,0.51989361702127657,0.3188087773390853,0.48010638297872338,0.20242306370714419,0.51989361702127657,0.3188087773390853,raw,-0.16553911820114384,0.46991567857167071,legacy 7-way line
+79,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + warmup=2000: MLM 0.1, LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_79_64bp_warmup_2000,247,47,200,0.48127659574468085,0.27255408190649444,0.51872340425531915,0.2279731614833477,0.51872340425531915,0.27255408190649444,flipped,-0.19781436852597997,0.4250841191515507,legacy 7-way line
+80,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.08: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_80_64bp_MLM_0.08,247,47,200,0.49787234042553191,0.22154087320831706,0.50212765957446803,0.24186517759529338,0.50212765957446803,0.24186517759529338,flipped,-0.22597313265086186,0.54483863098017571,legacy 7-way line
+81,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.12: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_81_64bp_MLM_0.12,247,47,200,0.50808510638297866,0.30715252370516188,0.49191489361702129,0.20628860995233253,0.50808510638297866,0.30715252370516188,raw,-0.11871336732316114,0.44459449327983186,legacy 7-way line
+82,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + MLM=0.12: cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_82_64bp_LR_0.00025_MLM_0.12,247,47,200,0.47808510638297874,0.20096873088093903,0.52191489361702126,0.22045434677637873,0.52191489361702126,0.22045434677637873,flipped,-0.24030197535449194,0.53019658815372761,legacy 7-way line
+83,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + warmup=500: MLM 0.1, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_83_64bp_LR_0.00025_warmup_500,247,47,200,0.47936170212765961,0.24085175111648066,0.52063829787234051,0.22224339310220231,0.52063829787234051,0.24085175111648066,flipped,-0.24546084688742634,0.43451126257827422,legacy 7-way line
+84,True,Hyperparameter Tuning: 64bp,"Exp 77优化 + 64bp窗口 + LR=0.00025 + warmup=2000: MLM 0.1, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_84_64bp_LR_0.00025_warmup_2000,247,47,200,0.4691489361702128,0.23400192109414786,0.5308510638297872,0.22552577558371206,0.5308510638297872,0.23400192109414786,flipped,-0.21777003593290384,0.47703118553690144,legacy 7-way line
+87,True,Hyperparameter Tuning: 64bp,"Exp 82优化 + 64bp窗口 + LR=0.00025 + MLM=0.12 + warmup=500: cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_87_64bp_LR_0.00025_MLM_0.12_warmup_500,247,47,200,0.43361702127659568,0.20761827502973879,0.56638297872340426,0.22851268654879955,0.56638297872340426,0.22851268654879955,flipped,-0.27408081582683302,0.5460025280741444,legacy 7-way line
+88,True,Hyperparameter Tuning: 64bp,"Exp 82优化 + 64bp窗口 + LR=0.00025 + MLM=0.12 + warmup=2000: cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_88_64bp_LR_0.00025_MLM_0.12_warmup_2000,247,47,200,0.44095744680851057,0.26881942297092182,0.55904255319148932,0.2452586045873652,0.55904255319148932,0.26881942297092182,flipped,-0.21636454248235293,0.43413595285581802,legacy 7-way line
+92,True,Multi-Seed Validation: 64bp,"Exp 80多种子验证 + 64bp窗口 + seed=123: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_92_64bp_MLM_0.08_seed123,247,47,200,0.52446808510638299,0.2735282074666297,0.47553191489361707,0.22351976815138527,0.52446808510638299,0.2735282074666297,raw,-0.20673360008942454,0.65575340245617397,legacy 7-way line
+93,True,Multi-Seed Validation: 64bp,"Exp 80多种子验证 + 64bp窗口 + seed=456: MLM 0.08, LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_93_64bp_MLM_0.08_seed456,247,47,200,0.39063829787234045,0.18165803156282823,0.60936170212765961,0.26450252302668109,0.60936170212765961,0.26450252302668109,flipped,-0.42074299268876975,0.5576566615938704,legacy 7-way line
+94,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + LR=0.00025: cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_94_64bp_MLM_0.08_LR_0.00025,247,47,200,0.4968085106382979,0.2951858808563787,0.5031914893617021,0.2039932319768763,0.5031914893617021,0.2951858808563787,flipped,-0.21296591459498232,0.47821205591076021,legacy 7-way line
+95,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + warmup=500: LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_95_64bp_MLM_0.08_warmup_500,247,47,200,0.48148936170212758,0.26052532694550834,0.51851063829787247,0.21033060445825513,0.51851063829787247,0.26052532694550834,flipped,-0.17555943095249685,0.43158171156435199,legacy 7-way line
+96,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.08 + warmup=2000: LR 0.0002, cosine, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_96_64bp_MLM_0.08_warmup_2000,247,47,200,0.49244680851063832,0.28437213744334949,0.50755319148936173,0.20946352132252397,0.50755319148936173,0.28437213744334949,flipped,-0.2204350411650623,0.43591014187610155,legacy 7-way line
+97,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.06: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_97_64bp_MLM_0.06,247,47,200,0.46638297872340423,0.25536254983543072,0.53361702127659572,0.22934561016409291,0.53361702127659572,0.25536254983543072,flipped,-0.18911182084064251,0.403247144312887,legacy 7-way line
+98,True,Hyperparameter Tuning: 64bp,"Exp 74优化 + 64bp窗口 + MLM=0.09: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_98_64bp_MLM_0.09,247,47,200,0.47946808510638295,0.23999987588753888,0.52053191489361705,0.20665982190104928,0.52053191489361705,0.23999987588753888,flipped,-0.22477992970933799,0.44774545273450572,legacy 7-way line
+99,True,Hyperparameter Tuning: 64bp,"Exp 80优化 + 64bp窗口 + MLM=0.07: LR 0.0002, cosine, warmup 1000, 512/8/6, n_aux=30, eval_steps=73 (1 epoch), batch=2048, max_position_embeddings=64",full_training_99_64bp_MLM_0.07,247,47,200,0.48372340425531912,0.27100288450392118,0.51627659574468088,0.21027361580809262,0.51627659574468088,0.27100288450392118,flipped,-0.22013148267259483,0.46443755446778917,legacy 7-way line
+,False,External,GPN-MSA trained checkpoint (not listed in base-model ablation registry),gpn_msa_trained,247,47,200,0.3645744680851064,0.14654578424985906,0.6354255319148936,0.25044838280677617,0.6354255319148936,0.25044838280677617,flipped,-2.4278762340545654,2.268017053604126,gpn-style checkpoint

thesis_repro/phylo_msa1_outputs/VEP_STRICTV2_MAIN_TEXT_MODELS_CI.csv

@@ -0,0 +1,15 @@
+model,model_display,n,pos,neg,polarity,auroc_cal,auroc_ci_low,auroc_ci_high,auprc_cal,auprc_ci_low,auprc_ci_high
+full_training_207,Exp. 207,2000,1000,1000,flipped,0.733943,0.7125272947801506,0.755020371391086,0.6847323524773955,0.6533980458282247,0.7186731616314354
+full_training_210,Exp. 210,2000,1000,1000,flipped,0.7262250000000001,0.7050979101994459,0.7468784719062471,0.6740753518571917,0.6423441740438786,0.7094963575058489
+full_training_206,Exp. 206,2000,1000,1000,flipped,0.725195,0.7032836364229388,0.7468506849830528,0.6691809398692765,0.6370310657718485,0.7031172905021881
+full_training_169,Exp. 169,2000,1000,1000,flipped,0.7239895000000001,0.7006639321766578,0.7444342453138575,0.6799283051027143,0.6489149925234776,0.7119381023560871
+full_training_209,Exp. 209,2000,1000,1000,flipped,0.718224,0.6962811138204397,0.7393275083430835,0.6662996670808599,0.6349055972837654,0.6999146079409072
+gpn_msa_trained,gpn\_msa\_trained,2000,1000,1000,flipped,0.692833,0.6699511182402567,0.7155528619662951,0.6579185849419642,0.627168339823677,0.6912906186345171
+full_training_201,Exp. 201,2000,1000,1000,flipped,0.6738085,0.650918558597175,0.6972209171517794,0.6293574033355777,0.5973498327311941,0.6633426161868698
+full_training_202,Exp. 202,2000,1000,1000,flipped,0.6654845,0.643022224664312,0.6887464733019331,0.624581599809271,0.5937271660108828,0.6591000393519586
+full_training_204,Exp. 204,2000,1000,1000,flipped,0.6625669999999999,0.6395363871218602,0.6864060297487499,0.6216540233650435,0.5907964290109966,0.6573355313579569
+full_training_203,Exp. 203,2000,1000,1000,flipped,0.6567280000000001,0.6338650048914087,0.6810401717928908,0.6173986577145278,0.5870013020276517,0.6521139293163132
+full_training_71_optimized_seed456,Exp. 71,2000,1000,1000,flipped,0.6519630000000001,0.6285995182116811,0.6752073919477776,0.6046712289973994,0.5736162644702323,0.638133640701809
+full_training_74_64bp,Exp. 74,2000,1000,1000,flipped,0.6453544999999999,0.6224719029412621,0.669327657617609,0.6076846745150857,0.5759849145623456,0.6416891469569679
+full_training_126,Exp. 126,2000,1000,1000,flipped,0.6186594999999999,0.5937146043120999,0.642589060474243,0.5665183401758486,0.5370781492289881,0.5988830226620359
+full_training_22,Exp. 22,2000,1000,1000,flipped,0.5726344999999999,0.5478138412415621,0.5978229370886969,0.5478864956412605,0.5179879904366504,0.5800305030361103

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_01.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:725ad694dc2baed44cb43002d489b2ecb6bbeccb6491e2e94b17929ee4b852ad
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_02.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3004ddfd45c1ccc66b6de74831fc648f4e9e9715d70f66c2265719573438843b
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_03.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14c34b54f358ae48453f69fd9014f69ddec21541dbf40a9115e77f433424a62a
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_04.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93988916483ff6c4dbc8de0585f0a01873246f2e83861cc5b9177d073bf71716
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_05.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19f08618f8fdb350a8b6e74b0ba5be3546c24ab1c6ebce45fa32d6a68b1691d3
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_06.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9630b6618bde062634924cf6d70a4277267373bcc5d0f2a8711296c0093e20b1
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_07.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44d0d92d4c84436e5ac76a33b072434e7f5d4a16c7741cbdc27a826bf90b4bc6
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_08.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:331ead69b8472cfecc67622a244bbaf5e0ea0d47acc201e625a19d5077206344
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_09.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46d39223cf5baf2c6aa181ec7aecf2b29de9e70f07b1bf239ac3630ef5eff3e1
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_10.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dad3f1aa5e68ab13e2aab5b80f13c9ac79e722a7a02d9d49f85f6f266f9c4bfd
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_100_64bp_MLM_0.085.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2496f5a06d8c1d99ec6a1f2b572e72dacf7cf27ff1fd420f17219bae8136afd2
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_101_64bp_LR_0.00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82963315576bc808539a0d795c346fe2ac06fffda7ffe894a373fc50cea8de6f
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_102_64bp_LR_0.00022.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d5e950fa119502e199a19a4b0a15d27fbfb0fb632c5de047ac610b89f9f8e8d
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_103_64bp_warmup_750.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:502c3b0fa2172db6be7bb00d0ad48aa3cddb92b1b7e81c3654017cdff587f289
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_104_64bp_warmup_1250.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f91c8780ba32290e9db0372ba78bba50060ddd6d8384036076dcea1c47a7698
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_105_64bp_warmup_1500.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc60e4a1e4be2e50f9413f8e72b098bd25b08425ebc1ffb08a7eaef086b7a942
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_106_64bp_soft_mask_0.15.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a8d407f42b7f0b4c3aaee6a7b0d2a14f65af0fd98301225c19592a5016d210f
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_107_64bp_class_weights.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:581625112868ec9e138b09030774f60091e50fbf68bd07c1731fddcddccc1f1f
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_108_64bp_focal_loss.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e43568065af533524234cdcbb7c55cc99e0eadcc47897300a223b07a408be925
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_109_64bp_768_12_8.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5deab9ee71528e06a04b6443b6a2a5c4e3f2342f35e419be33207782bde54073
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_11.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2992c2c7b1b0b73ee9029b14081ec7ff778edd0718a0488b2ca4f2b74567ccd
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_110_64bp_class_weights_aggressive.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c09755d141f93283a9407795ef586b00a7339948db6c5d20dc211487f5d2964
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_111_64bp_focal_loss_tuned.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fc5253d410ce5b1fd5ac856cbbf1aa634646f8f1cdb920d79ee96f4ef4272a9
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_112_64bp_combined_loss.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e5eb40bcb02e3f93417b663a9d659d53898b6bff44eb0241ca861f261c54011
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_113_64bp_focal_loss_gamma25.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c2dab38ab8a3216c66f0a0332068181fbbdafab73c1cd01afaef3192f211384
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_114_64bp_MLM_0.08_retrain.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c041877cccd2ab50ae7eb1c1448afbbf8a58c7e91af2cc2f1e9392ed6cd6ebe
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_115.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2d492e147997e58238e858111fc9f49d9de0c8f6a742b68983c139bbcf5e77
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_116.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a45db1092094895bf4d0ad795651d4913bdc06e3a1e7c3ff1910c3d612b5013
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_117.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2aba602c42ac21f887797fcacaf458afb27fb660123e93326ce2bd11689e6722
+size 2413

thesis_repro/phylo_msa1_outputs/vep_all_base_models/full_training_118.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6ee67e1a13b3802c1ff620460b18b12e28914ba144457a781c9016be8211e5c
+size 2413